From 81877145e1468ab4585687b15acf34f2a2c2b70d Mon Sep 17 00:00:00 2001
From: Maciej Wielgosz <maciej.wielgosz@nibio.no>
Date: Thu, 20 Apr 2023 13:37:23 +0200
Subject: [PATCH] update
---
cifar_example/cifar10_lightning.py | 113 +++++
cifar_example/cifar10_lightning_ver_2.py | 92 ++--
dgcnn/attention_usage.ipynb | 78 ++++
dgcnn/get_size_of_dataset.py | 6 +-
.../{ => jupyters}/edge_conv_layer_run.ipynb | 25 ++
dgcnn/jupyters/model10_vis.ipynb | 304 +++++++++++++
dgcnn/jupyters/my_shapenet_vis.ipynb | 414 ++++++++++++++++++
dgcnn/{ => jupyters}/transform_net_run.ipynb | 50 +++
dgcnn/my_models/model_shape_net.py | 4 +-
dgcnn/shapenet_data_dgcnn.py | 42 +-
10 files changed, 1064 insertions(+), 64 deletions(-)
create mode 100644 cifar_example/cifar10_lightning.py
create mode 100644 dgcnn/attention_usage.ipynb
rename dgcnn/{ => jupyters}/edge_conv_layer_run.ipynb (93%)
create mode 100644 dgcnn/jupyters/model10_vis.ipynb
create mode 100644 dgcnn/jupyters/my_shapenet_vis.ipynb
rename dgcnn/{ => jupyters}/transform_net_run.ipynb (65%)
diff --git a/cifar_example/cifar10_lightning.py b/cifar_example/cifar10_lightning.py
new file mode 100644
index 0000000..bec2a49
--- /dev/null
+++ b/cifar_example/cifar10_lightning.py
@@ -0,0 +1,113 @@
+import os
+from torch import nn
+import torch
+import torch.nn.functional as F
+from torch.utils.data import DataLoader
+from torchvision import transforms
+
+from pytorch_lightning import LightningModule, Trainer
+from torchvision.datasets import CIFAR10
+from pytorch_lightning.callbacks.progress import TQDMProgressBar
+from pytorch_lightning.loggers import CSVLogger
+
+# import modules
+from cifar_example.cifar_transformer_modules.my_transformer_layer import MyTransformerLayer
+from cifar_example.cifar_transformer_modules.embedding import Embedding
+
+# variables
+PATH_DATASETS = os.environ.get("PATH_DATASETS", ".")
+BATCH_SIZE = 256 if torch.cuda.is_available() else 64
+
+class CIFAR10LightningTransformer(LightningModule):
+ def __init__(self):
+ super().__init__()
+ self.embedding_size=64
+ self.criterion = torch.nn.CrossEntropyLoss()
+ self.embedding = Embedding(
+ patch_size=8,
+ in_channels=3,
+ out_channels=self.embedding_size,
+ return_patches=False,
+ extra_token=True
+ )
+ self.self_attention = MyTransformerLayer(d_model=self.embedding_size, nhead=16, dropout=0.3)
+ self.fc = nn.Linear(self.embedding_size, 10)
+
+ def forward(self, x):
+ embedding = self.embedding(x)
+ context = self.self_attention(embedding)
+ # get the first token
+ context = context[:, 0, :]
+
+ # context = context.mean(dim=1)
+
+ # get the classification
+ context = self.fc(context)
+
+ return context
+
+ def accuracy(self, logits, y):
+ preds = torch.argmax(logits, dim=1)
+ return torch.sum(preds == y).item() / len(y)
+
+ def training_step(self, batch, batch_idx):
+ x, y = batch
+ logits = self(x)
+ loss = self.criterion(logits, y)
+ self.log('train_loss', loss)
+ self.log('train_acc', self.accuracy(logits, y))
+ return loss
+
+ def validation_step(self, batch, batch_idx):
+ x, y = batch
+ logits = self(x)
+ loss = self.criterion(logits, y)
+ self.log('val_loss', loss)
+ self.log('val_acc', self.accuracy(logits, y))
+ return loss
+
+ def test_step(self, batch, batch_idx):
+ x, y = batch
+ logits = self(x)
+ loss = self.criterion(logits, y)
+ self.log('test_loss', loss)
+ self.log('test_acc', self.accuracy(logits, y))
+ return loss
+
+ def configure_optimizers(self):
+ return torch.optim.Adam(self.parameters(), lr=0.02)
+
+cifar_model = CIFAR10LightningTransformer()
+
+# train_ds = CIFAR10(PATH_DATASETS, train=True, download=True, transform=transforms.ToTensor())
+# train_loader = DataLoader(train_ds, batch_size=BATCH_SIZE)
+# val_ds = CIFAR10(PATH_DATASETS, train=False, download=True, transform=transforms.ToTensor())
+# val_loader = DataLoader(val_ds, batch_size=BATCH_SIZE)
+# test_ds = CIFAR10(PATH_DATASETS, train=False, download=True, transform=transforms.ToTensor())
+# test_loader = DataLoader(test_ds, batch_size=BATCH_SIZE)
+
+# get the train data
+train_ds = CIFAR10(root='./data', train=True, download=True, transform=transforms.ToTensor())
+# get test data
+test_ds = CIFAR10(root='./data', train=False, download=True, transform=transforms.ToTensor())
+
+# get the train loader
+train_loader = torch.utils.data.DataLoader(train_ds, batch_size=32, shuffle=True)
+
+# get the test loader
+test_loader = torch.utils.data.DataLoader(test_ds, batch_size=32, shuffle=False)
+
+# Initialize a trainer
+trainer = Trainer(
+ accelerator="auto",
+ devices=1 if torch.cuda.is_available() else None, # limiting got iPython runs
+ max_epochs=10,
+ callbacks=[TQDMProgressBar(refresh_rate=20)],
+ logger=CSVLogger(save_dir="logs/"),
+)
+
+# Train the model ⚡
+# trainer.fit(cifar_model, train_loader)
+
+# Test the model
+trainer.test(cifar_model, dataloaders=test_loader)
\ No newline at end of file
diff --git a/cifar_example/cifar10_lightning_ver_2.py b/cifar_example/cifar10_lightning_ver_2.py
index c12ca99..c23b1d4 100644
--- a/cifar_example/cifar10_lightning_ver_2.py
+++ b/cifar_example/cifar10_lightning_ver_2.py
@@ -54,57 +54,57 @@ from cifar_example.cifar_transformer_modules.my_transformer_layer import MyTrans
from cifar_example.cifar_transformer_modules.embedding import Embedding
# create resnet model for cifar10 classification
-class CIFAR10Model(pl.LightningModule):
- def __init__(self):
- super().__init__()
- self.model = self.create_model()
+# class CIFAR10Model(pl.LightningModule):
+# def __init__(self):
+# super().__init__()
+# self.model = self.create_model()
- def forward(self, x):
- return self.model(x)
+# def forward(self, x):
+# return self.model(x)
- def create_model(self):
- model = resnet18(pretrained=False, num_classes=10)
- model.conv1 = nn.Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
- model.maxpool = nn.Identity()
- return model
-
- def training_step(self, batch, batch_idx):
- x, y = batch
- y_hat = self(x)
- loss = F.cross_entropy(y_hat, y)
- self.log("train_loss", loss, on_step=True, on_epoch=True, prog_bar=True, logger=True)
- # log accuracy
- self.log("train_acc", self.accuracy(y_hat, y), on_step=True, on_epoch=True, prog_bar=True, logger=True)
- return loss
-
- def validation_step(self, batch, batch_idx):
- x, y = batch
- y_hat = self(x)
- loss = F.cross_entropy(y_hat, y)
- self.log("val_loss", loss, on_step=True, on_epoch=True, prog_bar=True, logger=True)
- # log accuracy
- self.log("val_acc", self.accuracy(y_hat, y), on_step=True, on_epoch=True, prog_bar=True, logger=True)
- return loss
-
- def test_step(self, batch, batch_idx):
- x, y = batch
- y_hat = self(x)
- loss = F.cross_entropy(y_hat, y)
- self.log("test_loss", loss, on_step=True, on_epoch=True, prog_bar=True, logger=True)
- # log accuracy
- self.log("test_acc", self.accuracy(y_hat, y), on_step=True, on_epoch=True, prog_bar=True, logger=True)
- return loss
-
- def configure_optimizers(self):
- return torch.optim.Adam(self.parameters(), lr=0.001)
-
- def accuracy(self, y_hat, y):
- preds = torch.argmax(y_hat, dim=1)
- return (preds == y).float().mean()
+# def create_model(self):
+# model = resnet18(pretrained=False, num_classes=10)
+# model.conv1 = nn.Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
+# model.maxpool = nn.Identity()
+# return model
+
+# def training_step(self, batch, batch_idx):
+# x, y = batch
+# y_hat = self(x)
+# loss = F.cross_entropy(y_hat, y)
+# self.log("train_loss", loss, on_step=True, on_epoch=True, prog_bar=True, logger=True)
+# # log accuracy
+# self.log("train_acc", self.accuracy(y_hat, y), on_step=True, on_epoch=True, prog_bar=True, logger=True)
+# return loss
+
+# def validation_step(self, batch, batch_idx):
+# x, y = batch
+# y_hat = self(x)
+# loss = F.cross_entropy(y_hat, y)
+# self.log("val_loss", loss, on_step=True, on_epoch=True, prog_bar=True, logger=True)
+# # log accuracy
+# self.log("val_acc", self.accuracy(y_hat, y), on_step=True, on_epoch=True, prog_bar=True, logger=True)
+# return loss
+
+# def test_step(self, batch, batch_idx):
+# x, y = batch
+# y_hat = self(x)
+# loss = F.cross_entropy(y_hat, y)
+# self.log("test_loss", loss, on_step=True, on_epoch=True, prog_bar=True, logger=True)
+# # log accuracy
+# self.log("test_acc", self.accuracy(y_hat, y), on_step=True, on_epoch=True, prog_bar=True, logger=True)
+# return loss
+
+# def configure_optimizers(self):
+# return torch.optim.Adam(self.parameters(), lr=0.001)
+
+# def accuracy(self, y_hat, y):
+# preds = torch.argmax(y_hat, dim=1)
+# return (preds == y).float().mean()
# create resnet model for cifar10 classification
-# class CIFAR10Model(pl.LightningModule):
+class CIFAR10Model(pl.LightningModule):
def __init__(self):
super().__init__()
self.embedding_size=64
diff --git a/dgcnn/attention_usage.ipynb b/dgcnn/attention_usage.ipynb
new file mode 100644
index 0000000..4c7e038
--- /dev/null
+++ b/dgcnn/attention_usage.ipynb
@@ -0,0 +1,78 @@
+{
+ "cells": [
+ {
+ "cell_type": "code",
+ "execution_count": 12,
+ "metadata": {},
+ "outputs": [
+ {
+ "ename": "AssertionError",
+ "evalue": "embed_dim must be divisible by num_heads",
+ "output_type": "error",
+ "traceback": [
+ "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
+ "\u001b[0;31mAssertionError\u001b[0m Traceback (most recent call last)",
+ "\u001b[1;32m/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb Cell 1\u001b[0m in \u001b[0;36m2\n\u001b[1;32m <a href='vscode-notebook-cell://ssh-remote%2Boracle_docker/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=24'>25</a>\u001b[0m \u001b[39m# show how to use self attention\u001b[39;00m\n\u001b[1;32m <a href='vscode-notebook-cell://ssh-remote%2Boracle_docker/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=25'>26</a>\u001b[0m x \u001b[39m=\u001b[39m torch\u001b[39m.\u001b[39mrandn(\u001b[39m2\u001b[39m, \u001b[39m1024\u001b[39m, \u001b[39m3\u001b[39m)\n\u001b[0;32m---> <a href='vscode-notebook-cell://ssh-remote%2Boracle_docker/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=26'>27</a>\u001b[0m self_attention \u001b[39m=\u001b[39m SelfAttention(\u001b[39m3\u001b[39;49m, \u001b[39m2\u001b[39;49m, \u001b[39m0.1\u001b[39;49m)\n\u001b[1;32m <a href='vscode-notebook-cell://ssh-remote%2Boracle_docker/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=27'>28</a>\u001b[0m out \u001b[39m=\u001b[39m self_attention(x)\n\u001b[1;32m <a href='vscode-notebook-cell://ssh-remote%2Boracle_docker/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=28'>29</a>\u001b[0m \u001b[39mprint\u001b[39m(out\u001b[39m.\u001b[39mshape)\n",
+ "\u001b[1;32m/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb Cell 1\u001b[0m in \u001b[0;36m1\n\u001b[1;32m <a href='vscode-notebook-cell://ssh-remote%2Boracle_docker/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=10'>11</a>\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mnum_heads \u001b[39m=\u001b[39m num_heads\n\u001b[1;32m <a href='vscode-notebook-cell://ssh-remote%2Boracle_docker/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=11'>12</a>\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mdropout \u001b[39m=\u001b[39m dropout\n\u001b[0;32m---> <a href='vscode-notebook-cell://ssh-remote%2Boracle_docker/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/attention_usage.ipynb#W0sdnNjb2RlLXJlbW90ZQ%3D%3D?line=12'>13</a>\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mself_attention \u001b[39m=\u001b[39m MultiheadAttention(in_channels, num_heads\u001b[39m=\u001b[39;49mnum_heads, dropout\u001b[39m=\u001b[39;49mdropout)\n",
+ "File \u001b[0;32m~/.local/lib/python3.8/site-packages/torch/nn/modules/activation.py:960\u001b[0m, in \u001b[0;36mMultiheadAttention.__init__\u001b[0;34m(self, embed_dim, num_heads, dropout, bias, add_bias_kv, add_zero_attn, kdim, vdim, batch_first, device, dtype)\u001b[0m\n\u001b[1;32m 958\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mbatch_first \u001b[39m=\u001b[39m batch_first\n\u001b[1;32m 959\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mhead_dim \u001b[39m=\u001b[39m embed_dim \u001b[39m/\u001b[39m\u001b[39m/\u001b[39m num_heads\n\u001b[0;32m--> 960\u001b[0m \u001b[39massert\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mhead_dim \u001b[39m*\u001b[39m num_heads \u001b[39m==\u001b[39m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39membed_dim, \u001b[39m\"\u001b[39m\u001b[39membed_dim must be divisible by num_heads\u001b[39m\u001b[39m\"\u001b[39m\n\u001b[1;32m 962\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mnot\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_qkv_same_embed_dim:\n\u001b[1;32m 963\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mq_proj_weight \u001b[39m=\u001b[39m Parameter(torch\u001b[39m.\u001b[39mempty((embed_dim, embed_dim), \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mfactory_kwargs))\n",
+ "\u001b[0;31mAssertionError\u001b[0m: embed_dim must be divisible by num_heads"
+ ]
+ }
+ ],
+ "source": [
+ "# import pytorch attention\n",
+ "from torch.nn import MultiheadAttention\n",
+ "from torch import nn\n",
+ "import torch\n",
+ "\n",
+ "# implement self attention\n",
+ "class SelfAttention(nn.Module):\n",
+ " def __init__(self, in_channels, num_heads, dropout):\n",
+ " super(SelfAttention, self).__init__()\n",
+ " self.in_channels = in_channels\n",
+ " self.num_heads = num_heads\n",
+ " self.dropout = dropout\n",
+ " self.self_attention = MultiheadAttention(in_channels, num_heads=num_heads, dropout=dropout)\n",
+ "\n",
+ " def forward(self, x):\n",
+ " batch_size = x.size(0)\n",
+ " num_points = x.size(2)\n",
+ " x = x.view(batch_size, -1, num_points)\n",
+ " x = x.permute(1, 0, 2)\n",
+ " out, attn = self.self_attention(x, x, x)\n",
+ " out = out.permute(1, 0, 2)\n",
+ " out = out.view(batch_size, -1, num_points)\n",
+ " return out\n",
+ "\n",
+ "# show how to use self attention\n",
+ "x = torch.randn(2, 1024, 3)\n",
+ "self_attention = SelfAttention(3, 3, 0.1)\n",
+ "out = self_attention(x)\n",
+ "print(out.shape)\n",
+ "\n"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.8.10"
+ },
+ "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
diff --git a/dgcnn/get_size_of_dataset.py b/dgcnn/get_size_of_dataset.py
index af7cf04..1ee6847 100644
--- a/dgcnn/get_size_of_dataset.py
+++ b/dgcnn/get_size_of_dataset.py
@@ -7,7 +7,7 @@ shapenet_data_train = ShapenetDataDgcnn(
return_cls_label=True,
small_data=False,
small_data_size=1000,
- just_four_classes=True,
+ num_classes=True,
split='train',
norm=True
)
@@ -18,7 +18,7 @@ shapenet_data_test = ShapenetDataDgcnn(
return_cls_label=True,
small_data=False,
small_data_size=1000,
- just_four_classes=True,
+ num_classes=True,
split='test',
norm=True
)
@@ -29,7 +29,7 @@ shapenet_data_val = ShapenetDataDgcnn(
return_cls_label=True,
small_data=False,
small_data_size=1000,
- just_four_classes=True,
+ num_classes=True,
split='val',
norm=True
)
diff --git a/dgcnn/edge_conv_layer_run.ipynb b/dgcnn/jupyters/edge_conv_layer_run.ipynb
similarity index 93%
rename from dgcnn/edge_conv_layer_run.ipynb
rename to dgcnn/jupyters/edge_conv_layer_run.ipynb
index b27286d..6792c92 100644
--- a/dgcnn/edge_conv_layer_run.ipynb
+++ b/dgcnn/jupyters/edge_conv_layer_run.ipynb
@@ -207,6 +207,31 @@
"\n",
"print(\"neighbors:\", neighbors.shape)"
]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 8,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "torch.Size([32, 3])\n"
+ ]
+ }
+ ],
+ "source": [
+ "import torch\n",
+ "\n",
+ "# Assuming you have a tensor 'x' with shape (8, 32, 3)\n",
+ "x = torch.randn(8, 32, 3)\n",
+ "\n",
+ "# Remove the batch dimension\n",
+ "x_no_batch = x[1,:,:]\n",
+ "\n",
+ "print(x_no_batch.shape)"
+ ]
}
],
"metadata": {
diff --git a/dgcnn/jupyters/model10_vis.ipynb b/dgcnn/jupyters/model10_vis.ipynb
new file mode 100644
index 0000000..5d5b91c
--- /dev/null
+++ b/dgcnn/jupyters/model10_vis.ipynb
@@ -0,0 +1,304 @@
+{
+ "cells": [
+ {
+ "cell_type": "code",
+ "execution_count": 6,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "1.13.1+cu117\n"
+ ]
+ }
+ ],
+ "source": [
+ "import os\n",
+ "import torch\n",
+ "os.environ['TORCH'] = torch.__version__\n",
+ "print(torch.__version__)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 7,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "from glob import glob\n",
+ "from PIL import Image\n",
+ "from tqdm.auto import tqdm\n",
+ "\n",
+ "import wandb\n",
+ "\n",
+ "import torch\n",
+ "import torch.nn.functional as F\n",
+ "\n",
+ "import numpy as np\n",
+ "import networkx as nx\n",
+ "import matplotlib.pyplot as plt\n",
+ "from pyvis.network import Network\n",
+ "from mpl_toolkits.mplot3d import Axes3D\n",
+ "\n",
+ "import torch_geometric.transforms as T\n",
+ "from torch_geometric.datasets import ModelNet\n",
+ "from torch_geometric.loader import DataLoader\n",
+ "from torch_geometric.utils import to_networkx\n",
+ "from torch_geometric.nn import knn_graph, radius_graph"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 8,
+ "metadata": {},
+ "outputs": [
+ {
+ "data": {
+ "text/html": [
+ "wandb version 0.14.2 is available! To upgrade, please run:\n",
+ " $ pip install wandb --upgrade"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Tracking run with wandb version 0.13.10"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Run data is saved locally in <code>/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/jupyters/wandb/run-20230414_091510-h2iclcgp</code>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Syncing run <strong><a href='https://wandb.ai/maciej-wielgosz-nibio/pyg-point-cloud/runs/h2iclcgp' target=\"_blank\">modelnet10/train/sampling-comparison</a></strong> to <a href='https://wandb.ai/maciej-wielgosz-nibio/pyg-point-cloud' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ " View project at <a href='https://wandb.ai/maciej-wielgosz-nibio/pyg-point-cloud' target=\"_blank\">https://wandb.ai/maciej-wielgosz-nibio/pyg-point-cloud</a>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ " View run at <a href='https://wandb.ai/maciej-wielgosz-nibio/pyg-point-cloud/runs/h2iclcgp' target=\"_blank\">https://wandb.ai/maciej-wielgosz-nibio/pyg-point-cloud/runs/h2iclcgp</a>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "wandb_project = \"pyg-point-cloud\" #@param {\"type\": \"string\"}\n",
+ "wandb_run_name = \"modelnet10/train/sampling-comparison\" #@param {\"type\": \"string\"}\n",
+ "\n",
+ "wandb.init(project=wandb_project, entity=\"maciej-wielgosz-nibio\", name=wandb_run_name, job_type=\"eda\")\n",
+ "\n",
+ "# Set experiment configs to be synced with wandb\n",
+ "config = wandb.config\n",
+ "config.display_sample = 2048 #@param {type:\"slider\", min:256, max:4096, step:16}\n",
+ "config.modelnet_dataset_alias = \"ModelNet10\" #@param [\"ModelNet10\", \"ModelNet40\"] {type:\"raw\"}\n",
+ "\n",
+ "# Classes for ModelNet10 and ModelNet40\n",
+ "categories = sorted([\n",
+ " x.split(os.sep)[-2]\n",
+ " for x in glob(os.path.join(\n",
+ " config.modelnet_dataset_alias, \"raw\", '*', ''\n",
+ " ))\n",
+ "])\n",
+ "\n",
+ "\n",
+ "config.categories = categories"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 9,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "pre_transform = T.NormalizeScale()\n",
+ "transform = T.SamplePoints(config.display_sample)\n",
+ "train_dataset = ModelNet(\n",
+ " root=config.modelnet_dataset_alias,\n",
+ " name=config.modelnet_dataset_alias[-2:],\n",
+ " train=True,\n",
+ " transform=transform,\n",
+ " pre_transform=pre_transform\n",
+ ")\n",
+ "val_dataset = ModelNet(\n",
+ " root=config.modelnet_dataset_alias,\n",
+ " name=config.modelnet_dataset_alias[-2:],\n",
+ " train=False,\n",
+ " transform=transform,\n",
+ " pre_transform=pre_transform\n",
+ ")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 10,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "100%|██████████| 20/20 [00:01<00:00, 11.47it/s]\n",
+ "100%|██████████| 100/100 [00:02<00:00, 36.10it/s]\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ "Waiting for W&B process to finish... <strong style=\"color:green\">(success).</strong>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ " View run <strong style=\"color:#cdcd00\">modelnet10/train/sampling-comparison</strong> at: <a href='https://wandb.ai/maciej-wielgosz-nibio/pyg-point-cloud/runs/h2iclcgp' target=\"_blank\">https://wandb.ai/maciej-wielgosz-nibio/pyg-point-cloud/runs/h2iclcgp</a><br/>Synced 5 W&B file(s), 3 media file(s), 103 artifact file(s) and 0 other file(s)"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Find logs at: <code>./wandb/run-20230414_091510-h2iclcgp/logs</code>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "table = wandb.Table(columns=[\"Model\", \"Class\", \"Split\"])\n",
+ "category_dict = {key: 0 for key in config.categories}\n",
+ "for idx in tqdm(range(len(train_dataset[:20]))):\n",
+ " point_cloud = wandb.Object3D(train_dataset[idx].pos.numpy())\n",
+ " category = config.categories[int(train_dataset[idx].y.item())]\n",
+ " category_dict[category] += 1\n",
+ " table.add_data(\n",
+ " point_cloud,\n",
+ " category,\n",
+ " \"Train\"\n",
+ " )\n",
+ "\n",
+ "data = [[key, category_dict[key]] for key in config.categories]\n",
+ "wandb.log({\n",
+ " f\"{config.modelnet_dataset_alias} Class-Frequency Distribution\" : wandb.plot.bar(\n",
+ " wandb.Table(data=data, columns = [\"Class\", \"Frequency\"]),\n",
+ " \"Class\", \"Frequency\",\n",
+ " title=f\"{config.modelnet_dataset_alias} Class-Frequency Distribution\"\n",
+ " )\n",
+ "})\n",
+ "\n",
+ "table = wandb.Table(columns=[\"Model\", \"Class\", \"Split\"])\n",
+ "category_dict = {key: 0 for key in config.categories}\n",
+ "for idx in tqdm(range(len(val_dataset[:100]))):\n",
+ " point_cloud = wandb.Object3D(val_dataset[idx].pos.numpy())\n",
+ " category = config.categories[int(val_dataset[idx].y.item())]\n",
+ " category_dict[category] += 1\n",
+ " table.add_data(\n",
+ " point_cloud,\n",
+ " category,\n",
+ " \"Test\"\n",
+ " )\n",
+ "wandb.log({config.modelnet_dataset_alias: table})\n",
+ "\n",
+ "data = [[key, category_dict[key]] for key in config.categories]\n",
+ "wandb.log({\n",
+ " f\"{config.modelnet_dataset_alias} Class-Frequency Distribution\" : wandb.plot.bar(\n",
+ " wandb.Table(data=data, columns = [\"Class\", \"Frequency\"]),\n",
+ " \"Class\", \"Frequency\",\n",
+ " title=f\"{config.modelnet_dataset_alias} Class-Frequency Distribution\"\n",
+ " )\n",
+ "})\n",
+ "\n",
+ "wandb.finish()"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": []
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.8.10"
+ },
+ "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
diff --git a/dgcnn/jupyters/my_shapenet_vis.ipynb b/dgcnn/jupyters/my_shapenet_vis.ipynb
new file mode 100644
index 0000000..bbd92c6
--- /dev/null
+++ b/dgcnn/jupyters/my_shapenet_vis.ipynb
@@ -0,0 +1,414 @@
+{
+ "cells": [
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import wandb\n",
+ "import random\n",
+ "random.seed(10)\n",
+ "\n",
+ "\n",
+ "wandb_project = \"my_shapenet_vis1\" \n",
+ "wandb_run_name = \"my_shapenet_vis_showcase\" \n",
+ "WANDB_NOTEBOOK_NAME = \"my_shapenet_vis.ipynb\"\n",
+ "\n",
+ "wandb.init(project=wandb_project, entity=\"maciej-wielgosz-nibio\", name=wandb_run_name, job_type=\"eda\")\n",
+ "\n",
+ "# load shape net data\n",
+ "import sys\n",
+ "sys.path.append('/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn')\n",
+ "from shapenet_data_dgcnn import ShapenetDataDgcnn\n",
+ "shapenet_data = ShapenetDataDgcnn(\n",
+ " root='/home/nibio/mutable-outside-world/code/oracle_gpu_runs/data/shapenet',\n",
+ " npoints=512,\n",
+ " return_cls_label=True,\n",
+ " small_data=True,\n",
+ " small_data_size=10,\n",
+ " num_classes=1,\n",
+ " data_augmentation=False,\n",
+ " split='train',\n",
+ " norm=True\n",
+ " )\n",
+ "\n",
+ "# get first data point\n",
+ "data = shapenet_data[1]\n",
+ "data[0]\n",
+ "\n",
+ "# print(data[1])\n",
+ "\n",
+ "# find how many different values are in data[1]\n",
+ "import numpy as np\n",
+ "uv = np.unique(data[1])\n",
+ "\n",
+ "# generte random RGB colors for each class\n",
+ "import random\n",
+ "colors = []\n",
+ "for i in range(50):\n",
+ " colors.append([random.randint(0, 255), random.randint(0, 255), random.randint(0, 255)])\n",
+ "\n",
+ "colors_per_point = [colors[data[1][i]] for i in range(len(data[1]))]\n",
+ "\n",
+ "# create a point cloud\n",
+ "points_rgb = np.array([[p[0], p[1], p[2], c[0], c[1], c[2]] for p, c in zip(data[0], colors_per_point)])\n",
+ "\n",
+ "point_cloud = wandb.Object3D(\n",
+ " {\n",
+ " \"type\": \"lidar/beta\",\n",
+ " \"points\": points_rgb\n",
+ " }\n",
+ ")\n",
+ "\n",
+ "# show point cloud in wandb\n",
+ "wandb.log({\"point_cloud\": point_cloud})\n",
+ "\n",
+ "\n",
+ "# create a point cloud points_rgb_0 by choosing only points with class 0\n",
+ "points_rgb_0 = points_rgb[data[1] == 0]\n",
+ "\n",
+ "print(points_rgb_0.shape)\n",
+ "\n",
+ "point_cloud_0 = wandb.Object3D(\n",
+ " {\n",
+ " \"type\": \"lidar/beta\",\n",
+ " \"points\": points_rgb_0\n",
+ " }\n",
+ ")\n",
+ "\n",
+ "# create a point cloud points_rgb_1 by choosing only points with class 1\n",
+ "points_rgb_1 = points_rgb[data[1] == 1]\n",
+ "\n",
+ "print(points_rgb_1.shape)\n",
+ "\n",
+ "point_cloud_1 = wandb.Object3D(\n",
+ " {\n",
+ " \n",
+ " \"type\": \"lidar/beta\",\n",
+ " \"points\": points_rgb_1\n",
+ " }\n",
+ ")\n",
+ "\n",
+ "\n",
+ "# show point cloud in wandb\n",
+ "wandb.log({\"point_cloud_0\": point_cloud_0})\n",
+ "wandb.log({\"point_cloud_1\": point_cloud_1})\n",
+ "\n",
+ "# get the histogram for the data[1] (class labels)\n",
+ "labels_hist = {}\n",
+ "for label in data[1]:\n",
+ " if label in labels_hist:\n",
+ " labels_hist[label] += 1\n",
+ " else:\n",
+ " labels_hist[label] = 1\n",
+ "\n",
+ "# create a table for the histogram\n",
+ "table = wandb.Table(columns=[\"Class\", \"frequency\"])\n",
+ "\n",
+ "# write the class histogram to wandb\n",
+ "for class_name, count in labels_hist.items():\n",
+ " table.add_data(class_name, count)\n",
+ "\n",
+ "wandb.log({\"class freq\": wandb.plot.bar(table, \"Class\", \"frequency\")})\n",
+ "\n",
+ "wandb.finish()\n",
+ "\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "metadata": {},
+ "outputs": [
+ {
+ "data": {
+ "text/html": [
+ "Finishing last run (ID:vrw80dza) before initializing another..."
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Waiting for W&B process to finish... <strong style=\"color:green\">(success).</strong>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ " View run <strong style=\"color:#cdcd00\">my_shapenet_vis_showcase</strong> at: <a href='https://wandb.ai/maciej-wielgosz-nibio/my_shapenet_vis1/runs/vrw80dza' target=\"_blank\">https://wandb.ai/maciej-wielgosz-nibio/my_shapenet_vis1/runs/vrw80dza</a><br/>Synced 5 W&B file(s), 1 media file(s), 31 artifact file(s) and 0 other file(s)"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Find logs at: <code>./wandb/run-20230417_115825-vrw80dza/logs</code>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Successfully finished last run (ID:vrw80dza). Initializing new run:<br/>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Tracking run with wandb version 0.14.2"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Run data is saved locally in <code>/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn/jupyters/wandb/run-20230417_120056-f4w7yi7s</code>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ "Syncing run <strong><a href='https://wandb.ai/maciej-wielgosz-nibio/my_shapenet_vis1/runs/f4w7yi7s' target=\"_blank\">my_shapenet_vis_showcase</a></strong> to <a href='https://wandb.ai/maciej-wielgosz-nibio/my_shapenet_vis1' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ " View project at <a href='https://wandb.ai/maciej-wielgosz-nibio/my_shapenet_vis1' target=\"_blank\">https://wandb.ai/maciej-wielgosz-nibio/my_shapenet_vis1</a>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/html": [
+ " View run at <a href='https://wandb.ai/maciej-wielgosz-nibio/my_shapenet_vis1/runs/f4w7yi7s' target=\"_blank\">https://wandb.ai/maciej-wielgosz-nibio/my_shapenet_vis1/runs/f4w7yi7s</a>"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ },
+ {
+ "data": {
+ "text/plain": [
+ "({0: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d575ea30>,\n",
+ " 1: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd857cd0>,\n",
+ " 2: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd91c8e0>,\n",
+ " 3: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd91caf0>,\n",
+ " 4: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d57509a0>,\n",
+ " 5: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d578bdf0>,\n",
+ " 6: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d36b7d30>,\n",
+ " 7: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d57fdca0>,\n",
+ " 8: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d57d6a60>,\n",
+ " 9: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd7e4250>},\n",
+ " {0: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd7e4490>,\n",
+ " 1: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd7dbb50>,\n",
+ " 2: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d57f70a0>,\n",
+ " 3: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d5789340>,\n",
+ " 4: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d57894c0>,\n",
+ " 5: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d5777310>,\n",
+ " 6: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d58012b0>,\n",
+ " 7: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d575e910>,\n",
+ " 8: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d5790040>,\n",
+ " 9: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd84a5e0>},\n",
+ " {0: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd84a700>,\n",
+ " 1: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd857eb0>,\n",
+ " 2: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1dd857e80>,\n",
+ " 3: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d5750d60>,\n",
+ " 4: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d5750f40>,\n",
+ " 5: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d5750820>,\n",
+ " 6: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d5750c70>,\n",
+ " 7: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d5750cd0>,\n",
+ " 8: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d578bfa0>,\n",
+ " 9: <wandb.sdk.data_types.object_3d.Object3D at 0x7fd1d578b700>})"
+ ]
+ },
+ "execution_count": 2,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "import wandb\n",
+ "import random\n",
+ "import numpy as np\n",
+ "random.seed(10)\n",
+ "\n",
+ "\n",
+ "wandb_project = \"my_shapenet_vis1\" \n",
+ "wandb_run_name = \"my_shapenet_vis_showcase\" \n",
+ "WANDB_NOTEBOOK_NAME = \"my_shapenet_vis.ipynb\"\n",
+ "\n",
+ "run = wandb.init(project=wandb_project, entity=\"maciej-wielgosz-nibio\", name=wandb_run_name, job_type=\"eda\")\n",
+ "\n",
+ "# load shape net data\n",
+ "import sys\n",
+ "sys.path.append('/home/nibio/mutable-outside-world/code/oracle_gpu_runs/dgcnn')\n",
+ "from shapenet_data_dgcnn import ShapenetDataDgcnn\n",
+ "shapenet_data = ShapenetDataDgcnn(\n",
+ " root='/home/nibio/mutable-outside-world/code/oracle_gpu_runs/data/shapenet',\n",
+ " npoints=512,\n",
+ " return_cls_label=True,\n",
+ " small_data=True,\n",
+ " small_data_size=10,\n",
+ " num_classes=1,\n",
+ " data_augmentation=False,\n",
+ " split='train',\n",
+ " norm=True\n",
+ " )\n",
+ "\n",
+ "# create a function which will create a point cloud for each class\n",
+ "def create_point_clouds(data, pred):\n",
+ " data_point_clouds = {}\n",
+ " pred_point_clouds = {}\n",
+ " diff_point_clouds = {}\n",
+ "\n",
+ " colors = []\n",
+ " for i in range(50):\n",
+ " colors.append([random.randint(0, 255), random.randint(0, 255), random.randint(0, 255)])\n",
+ "\n",
+ " for i in range(len(data)):\n",
+ " colors_per_point = [colors[data[i][1][j]] for j in range(len(data[i][1]))]\n",
+ " points_rgb = np.array([[p[0], p[1], p[2], c[0], c[1], c[2]] for p, c in zip(data[i][0], colors_per_point)])\n",
+ " data_point_clouds[i] = wandb.Object3D(\n",
+ " {\n",
+ " \"type\": \"lidar/beta\",\n",
+ " \"points\": points_rgb\n",
+ " }\n",
+ " )\n",
+ "\n",
+ " for i in range(len(pred)):\n",
+ " colors_per_point = [colors[pred[i][1][j]] for j in range(len(pred[i][1]))]\n",
+ " points_rgb = np.array([[p[0], p[1], p[2], c[0], c[1], c[2]] for p, c in zip(pred[i][0], colors_per_point)])\n",
+ " pred_point_clouds[i] = wandb.Object3D(\n",
+ " {\n",
+ " \"type\": \"lidar/beta\",\n",
+ " \"points\": points_rgb\n",
+ " }\n",
+ " )\n",
+ " \n",
+ " for i in range(len(data)):\n",
+ " diff_point_clouds[i] = data[i][0] - pred[i][0] + 1\n",
+ " diff_point_clouds[i] = wandb.Object3D(\n",
+ " {\n",
+ " \"type\": \"lidar/beta\",\n",
+ " \"points\": diff_point_clouds[i],\n",
+ " \n",
+ " }\n",
+ " )\n",
+ " \n",
+ " # wandb.log({\"point_cloud_\" + str(i): point_clouds[i]})\n",
+ " # table.add_data(point_clouds[i], point_clouds[i])\n",
+ " table_data = [[data_point_clouds[i], pred_point_clouds[i], diff_point_clouds[i]] for i in range(len(data_point_clouds))]\n",
+ " table = wandb.Table(data=table_data, columns=[\"gt\", \"pred\", \"diff\"])\n",
+ "\n",
+ " # show the table in wandb\n",
+ " run.log({\"point_clouds table \": table})\n",
+ "\n",
+ " return data_point_clouds, pred_point_clouds, diff_point_clouds\n",
+ "\n",
+ "\n",
+ "# create a function which will create a histogram for each class\n",
+ "def create_histograms(data):\n",
+ " labels_hist = {}\n",
+ " for i in range(len(data)):\n",
+ " labels_hist[i] = {}\n",
+ " for label in data[i][1]:\n",
+ " if label in labels_hist[i]:\n",
+ " labels_hist[i][label] += 1\n",
+ " else:\n",
+ " labels_hist[i][label] = 1\n",
+ " \n",
+ "\n",
+ " # create a table for the histogram\n",
+ " table = wandb.Table(columns=[\"Class\", \"frequency\"])\n",
+ "\n",
+ " # write the class histogram to wandb\n",
+ " for class_name, count in labels_hist.items():\n",
+ " table.add_data(class_name, count)\n",
+ "\n",
+ " wandb.log({\"class freq\": wandb.plot.bar(table, \"Class\", \"frequency\")})\n",
+ " \n",
+ "\n",
+ "# use the functions to create point clouds and histograms\n",
+ "create_point_clouds(shapenet_data, shapenet_data)\n",
+ "\n"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3",
+ "language": "python",
+ "name": "python3"
+ },
+ "language_info": {
+ "codemirror_mode": {
+ "name": "ipython",
+ "version": 3
+ },
+ "file_extension": ".py",
+ "mimetype": "text/x-python",
+ "name": "python",
+ "nbconvert_exporter": "python",
+ "pygments_lexer": "ipython3",
+ "version": "3.8.10"
+ },
+ "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
diff --git a/dgcnn/transform_net_run.ipynb b/dgcnn/jupyters/transform_net_run.ipynb
similarity index 65%
rename from dgcnn/transform_net_run.ipynb
rename to dgcnn/jupyters/transform_net_run.ipynb
index fa3166b..213d882 100644
--- a/dgcnn/transform_net_run.ipynb
+++ b/dgcnn/jupyters/transform_net_run.ipynb
@@ -81,6 +81,56 @@
"# Apply the transformation matrix to the input point cloud\n",
"input_tensor_transformed = torch.bmm(transform_matrix, input_tensor)\n"
]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 14,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Ground truth: tensor([[1, 0],\n",
+ " [0, 1],\n",
+ " [1, 0],\n",
+ " [0, 1],\n",
+ " [1, 0],\n",
+ " [0, 1]])\n",
+ "Predicted: tensor([[1, 0],\n",
+ " [0, 1],\n",
+ " [0, 1],\n",
+ " [0, 1],\n",
+ " [1, 0],\n",
+ " [0, 1]])\n",
+ "Jaccard index: tensor(0.7500)\n"
+ ]
+ }
+ ],
+ "source": [
+ "import torch\n",
+ "import torchmetrics\n",
+ "\n",
+ "# Define the Jaccard index (IoU) metric\n",
+ "jaccard_index = torchmetrics.JaccardIndex(num_classes=2, average=\"macro\", task=\"binary\")\n",
+ "\n",
+ "# Example ground truth and predicted labels\n",
+ "ground_truth = torch.tensor([0, 1, 0, 1, 0, 1], dtype=torch.long)\n",
+ "predicted = torch.tensor([0, 1, 1, 1, 0, 1], dtype=torch.long)\n",
+ "\n",
+ "# One-hot encode the ground truth and predicted labels\n",
+ "ground_truth_one_hot = torch.nn.functional.one_hot(ground_truth, num_classes=2)\n",
+ "predicted_one_hot = torch.nn.functional.one_hot(predicted, num_classes=2)\n",
+ "\n",
+ "print(\"Ground truth: \", ground_truth_one_hot)\n",
+ "print(\"Predicted: \", predicted_one_hot)\n",
+ "\n",
+ "\n",
+ "# Compute the Jaccard index\n",
+ "# jaccard_index = jaccard_index(predicted_one_hot, ground_truth_one_hot)\n",
+ "jaccard_index = jaccard_index(predicted, ground_truth)\n",
+ "print(\"Jaccard index: \", jaccard_index)"
+ ]
}
],
"metadata": {
diff --git a/dgcnn/my_models/model_shape_net.py b/dgcnn/my_models/model_shape_net.py
index 3593570..fc275ac 100644
--- a/dgcnn/my_models/model_shape_net.py
+++ b/dgcnn/my_models/model_shape_net.py
@@ -7,7 +7,7 @@ from my_models.edge_conv_new import EdgeConvNew
class DgcnShapeNet(nn.Module):
- def __init__(self, seg_num_all):
+ def __init__(self, seg_num_all, num_classes):
super(DgcnShapeNet, self).__init__()
self.seg_num_all = seg_num_all
self.transform_net = Transform_Net()
@@ -25,7 +25,7 @@ class DgcnShapeNet(nn.Module):
self.conv6 = nn.Sequential(nn.Conv1d(256, 256, kernel_size=1, bias=False),
self.bn6,
nn.LeakyReLU(negative_slope=0.2))
- self.conv7 = nn.Sequential(nn.Conv1d(16, 64, kernel_size=1, bias=False),
+ self.conv7 = nn.Sequential(nn.Conv1d(num_classes, 64, kernel_size=1, bias=False),
self.bn7,
nn.LeakyReLU(negative_slope=0.2))
self.conv8 = nn.Sequential(nn.Conv1d(512, 256, kernel_size=1, bias=False),
diff --git a/dgcnn/shapenet_data_dgcnn.py b/dgcnn/shapenet_data_dgcnn.py
index abf9ba9..d518a63 100644
--- a/dgcnn/shapenet_data_dgcnn.py
+++ b/dgcnn/shapenet_data_dgcnn.py
@@ -17,7 +17,7 @@ class ShapenetDataDgcnn(object):
small_data=False,
small_data_size=10,
return_cls_label=False,
- just_four_classes=False,
+ num_classes=1, # None - all classes (50), 1 - one class, 2 - two classes, max 4
norm=False,
augmnetation=False,
data_augmentation=False
@@ -29,7 +29,7 @@ class ShapenetDataDgcnn(object):
self.small_data = small_data
self.small_data_size = small_data_size
self.return_cls_label = return_cls_label
- self.just_four_classes = just_four_classes
+ self.num_classes = num_classes
self.norm = norm
self.augmnetation = augmnetation
self.data_augmentation = data_augmentation
@@ -90,12 +90,28 @@ class ShapenetDataDgcnn(object):
# get one class of data
# get the the number of the class airplane
- if self.just_four_classes:
- out_data = [x for x in out_data if x.split('/')[-2] in [
+ if self.num_classes is not None:
+ if self.num_classes == 1:
+ out_data = [x for x in out_data if x.split('/')[-2] in [
+ self.cat['Airplane']
+ ]]
+ elif self.num_classes == 2:
+ out_data = [x for x in out_data if x.split('/')[-2] in [
+ self.cat['Airplane'],
+ self.cat['Lamp']
+ ]]
+ elif self.num_classes == 3:
+ out_data = [x for x in out_data if x.split('/')[-2] in [
+ self.cat['Airplane'],
+ self.cat['Lamp'],
+ self.cat['Car']
+ ]]
+ elif self.num_classes == 4:
+ out_data = [x for x in out_data if x.split('/')[-2] in [
self.cat['Airplane'],
self.cat['Lamp'],
- self.cat['Chair'],
- self.cat['Table'],
+ self.cat['Car'],
+ self.cat['Chair']
]]
return out_data
@@ -207,8 +223,6 @@ class ShapenetDataDgcnn(object):
point_set = self.rotate_pointcloud(point_set)
point_set = self.translate_pointcloud(point_set)
-
-
choice = np.random.choice(len(point_set), self.npoints, replace=True)
point_set = point_set[choice, :]
@@ -227,16 +241,18 @@ class ShapenetDataDgcnn(object):
class_name = self.val_data_file[index].split('/')[-2]
# apply the mapper
- if self.just_four_classes:
- class_name = self.class_mapper_4_classes(class_name)
- else:
- class_name = self.class_mapper(class_name)
+ # if self.num_classes:
+ # class_name = self.class_mapper_4_classes(class_name)
+ # else:
+ # class_name = self.class_mapper(class_name)
+ class_name = self.class_mapper(class_name)
+
# convert the class name to a number
class_name = np.array(class_name, dtype=np.int64)
# map to tensor
- class_name = torch.from_numpy(class_name)
+ # class_name = torch.from_numpy(class_name)
if self.return_cls_label:
return point_set, labels, class_name
--
GitLab