Merge pull request #325 from Visual-Behavior/bv0.5.0-beta

Bv0.5.0 beta

Author: thibo73800

Dockerfile

@@ -1,17 +1,37 @@
 # tagged aloception-oss:cuda-11.3.1-pytorch1.10.1-lightning1.4.1
+
 FROM nvidia/cuda:11.3.1-cudnn8-devel-ubuntu20.04
+#FROM nvidia/cuda:11.6.0-cudnn8-devel-ubuntu20.04
+
+ARG py=3.9
+ARG pytorch=1.13.1
+ARG torchvision=0.14.1
+ARG torchaudio=0.13.1
+ARG pytorch_lightning=1.9.0
+ARG pycyda=11.7
+
 ENV TZ=Europe/Paris
 ENV DEBIAN_FRONTEND=noninteractive
+
 RUN apt-get update
 RUN apt-get install -y build-essential nano git wget libgl1-mesa-glx
+
 # Usefull for scipy 
 RUN apt-get install -y gfortran 
+# required for aloscene
+RUN apt-get install -y libglib2.0-0
+
+
 RUN wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
 RUN bash Miniconda3-latest-Linux-x86_64.sh -b -p /miniconda
 ENV PATH=$PATH:/miniconda/condabin:/miniconda/bin
 RUN /bin/bash -c "source activate base"
 ENV HOME /workspace
 WORKDIR /workspace
-RUN conda install python=3.9 pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=11.3 opencv=4.5.3 -c pytorch -c conda-forge
-COPY requirements.txt /install/requirements.txt
-RUN pip install -r /install/requirements.txt
+
+# Pytorch & pytorch litning
+RUN conda install pytorch==${pytorch} torchvision==${torchvision} torchaudio==${torchaudio} pytorch-cuda=${pycuda} -c pytorch -c nvidia
+RUN pip install pytorch_lightning==${pytorch_lightning}
+
+COPY requirements-torch1.13.1.txt /install/requirements-torch1.13.1.txt
+RUN pip install -r /install/requirements-torch1.13.1.txt

README.md

@@ -1,10 +1,10 @@
 <p align="center">
-  <img src="images/aloception.png" style="text-align:center; width: 50%;" alt="Logo aloception" />
+  <img src="images/aloception-oss.jpg" style="text-align:center; width: 50%;" alt="Logo aloception" />
 </p>
 
 <a href="https://visual-behavior.github.io/aloception-oss/">Documentation</a>
 
-[![Conventional Commits](https://img.shields.io/badge/Conventional%20Commits-0.3.0-green.svg)](https://conventionalcommits.org)
+[![Conventional Commits](https://img.shields.io/badge/Conventional%20Commits-0.5.0-green.svg)](https://conventionalcommits.org)
 
 # Aloception open source software
 
@@ -75,6 +75,26 @@ training pipelines with **augmented tensors**.
 
 ## Installation
 
+### Docker install
+
+```
+docker build -t aloception-oss:cuda-11.3.1-pytorch1.13.1-lightning1.9.0 .
+```
+
+```
+docker run --gpus all -it -v /YOUR/WORKSPACE/:/workspace --privileged -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix aloception-oss:cuda-11.3.1-pytorch1.13.1-lightning1.9.0
+```
+
+Or without building the image
+
+```
+docker run --gpus all -it -v /YOUR/WORKSPACE/:/workspace --privileged -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix visualbehaviorofficial/aloception-oss:cuda-11.3.1-pytorch1.13.1-lightning1.9.0
+```
+
+
+
+### Pip install
+
 You first need to install PyTorch 1.10.1 based on your hardware and environment
 configuration. Please refer to the [pytorch website](https://pytorch.org/get-started/locally/) for this installation.
 
@@ -90,7 +110,7 @@ Alternatively, you can clone the repository and use:
 pip install -e aloception-oss/
 ```
 
-Or setup the repo yourself in your env and install the dependencies 
+Or setup the repo yourself in your env and install the dependencies
 
 ```sh
 pip install -r requirements.txt

alodataset/__init__.py

@@ -14,5 +14,6 @@
 from .sintel_flow_dataset import SintelFlowDataset
 from .sintel_disparity_dataset import SintelDisparityDataset
 from .sintel_multi_dataset import SintelMultiDataset
+from .from_directory_dataset import FromDirectoryDataset
 from .woodScape_dataset import WooodScapeDataset
 from .woodScape_split_dataset import WoodScapeSplitDataset

alodataset/base_dataset.py

@@ -56,7 +56,7 @@ def stream_loader(dataset, num_workers=2):
     return data_loader
 
 
-def train_loader(dataset, batch_size=1, num_workers=2, sampler=torch.utils.data.RandomSampler):
+def train_loader(dataset, batch_size=1, num_workers=2, sampler=torch.utils.data.RandomSampler, sampler_kwargs={}):
     """Get training loader from the dataset
 
     Parameters
@@ -69,14 +69,15 @@ def train_loader(dataset, batch_size=1, num_workers=2, sampler=torch.utils.data.
         Number of workers, by default 2
     sampler : torch.utils.data, optional
         Callback to sampler the dataset, by default torch.utils.data.RandomSampler
+        Or instance of any class inheriting from torch.utils.data.Sampler
 
     Returns
     -------
     torch.utils.data.DataLoader
         A generator
     """
-    sampler = sampler(dataset) if sampler is not None else None
-
+    if sampler is not None and not(isinstance(sampler, torch.utils.data.Sampler)):
+        sampler = sampler(dataset, **sampler_kwargs)
     data_loader = torch.utils.data.DataLoader(
         dataset,
         # batch_sampler=batch_sampler,
@@ -332,7 +333,7 @@ def stream_loader(self, num_workers=2):
         """
         return stream_loader(self, num_workers=num_workers)
 
-    def train_loader(self, batch_size=1, num_workers=2, sampler=torch.utils.data.RandomSampler):
+    def train_loader(self, batch_size=1, num_workers=2, sampler=torch.utils.data.RandomSampler, sampler_kwargs={}):
         """Get training loader from the dataset
 
         Parameters
@@ -351,7 +352,7 @@ def train_loader(self, batch_size=1, num_workers=2, sampler=torch.utils.data.Ran
         torch.utils.data.DataLoader
             A generator
         """
-        return train_loader(self, batch_size=batch_size, num_workers=num_workers, sampler=sampler)
+        return train_loader(self, batch_size=batch_size, num_workers=num_workers, sampler=sampler, sampler_kwargs=sampler_kwargs    )
 
     def prepare(self):
         """Prepare the dataset. Not all child class need to implement this method.

alodataset/coco_base_dataset.py

@@ -6,13 +6,15 @@
 import os
 import numpy as np
 import torch
+
+from alodataset import BaseDataset
+from aloscene import BoundingBoxes2D, Frame, Labels, Mask
 from collections import defaultdict
+from pathlib import Path
 from pycocotools import mask as coco_mask
 from pycocotools.coco import COCO
 from typing import Dict, Union
 
-from alodataset import BaseDataset
-from aloscene import BoundingBoxes2D, Frame, Labels, Mask
 
 
 class CocoBaseDataset(BaseDataset):
@@ -78,10 +80,17 @@ def __init__(
             return
         else:
             assert img_folder is not None, "When sample = False, img_folder must be given."
-            assert ann_file is not None, "When sample = False, ann_file must be given."
+            assert ann_file is not None or "test" in img_folder, "When sample = False and the test split is not used, ann_file must be given."
+
 
         # Create properties
         self.img_folder = os.path.join(self.dataset_dir, img_folder)
+
+        if "test" in img_folder:
+            #get a list of  indices that don't rely on the annotation file
+            self.items = [int(Path(os.path.join(self.img_folder, f)).stem) for f in os.listdir(self.img_folder) if os.path.isfile(os.path.join(self.img_folder, f))]
+            return
+
         self.coco = COCO(os.path.join(self.dataset_dir, ann_file))
         self.items = list(sorted(self.coco.imgs.keys()))
 
@@ -231,7 +240,12 @@ def getitem(self, idx):
             return BaseDataset.__getitem__(self, idx)
 
         image_id = self.items[idx]
+        if "test" in self.img_folder:
+            #get the filename from image_id without relying on annotation file
+            return Frame(os.path.join(self.img_folder, f"{str(image_id).zfill(12)}.jpg"))
+
         frame = Frame(os.path.join(self.img_folder, self.coco.loadImgs(image_id)[0]["file_name"]))
+
         target = self.coco.loadAnns(self.coco.getAnnIds(image_id))
         target = {"image_id": image_id, "annotations": target}
         _, target = self.prepare(frame, target)
@@ -341,7 +355,12 @@ def __call__(self, image, target):
 
 
 if __name__ == "__main__":
-    coco_dataset = CocoBaseDataset(sample=True)
+    coco_dataset = CocoBaseDataset(sample=False, img_folder="test2017")
+    #checking if regular getitem works
+    frame = coco_dataset[0]
+    frame.get_view().render()
+
+    #check if dataloader works
     for f, frames in enumerate(coco_dataset.train_loader(batch_size=2)):
         frames = Frame.batch_list(frames)
         frames.get_view().render()

alodataset/crowd_human_dataset.py

@@ -36,7 +36,18 @@ def __init__(
             return
         else:
             assert img_folder is not None, "When sample = False, img_folder must be given."
-            assert ann_file is not None, "When sample = False, ann_file must be given."
+            assert ann_file is not None or "test" in img_folder, "When sample = False and the test split is not used, ann_file must be given."
+
+        if "test" in img_folder:
+            self._img_folder = img_folder
+            self.img_folder = os.path.join(self.dataset_dir, img_folder, "images_test")
+
+            self.items = []
+            for f in os.listdir(self.img_folder):
+                if os.path.isfile(os.path.join(self.img_folder, f)):
+                    self.items.append({"ID": Path(os.path.join(self.img_folder, f)).stem})
+
+            return
 
         assert type(img_folder) == type(ann_file), "img_folder & ann_file must be the same type."
 
@@ -121,9 +132,14 @@ def getitem(self, idx):
             return BaseDataset.__getitem__(self, idx)
 
         record = self.items[idx]
-        ann_id = record["ann_id"]
         image_id = record["ID"]
 
+        if "test" in self.img_folder:
+            #get the filename from image_id without relying on annotation file
+            return Frame(os.path.join(self.img_folder, image_id + ".jpg"))
+
+        ann_id = record["ann_id"]
+
         image_path = os.path.join(self.img_folder[ann_id], image_id + ".jpg")
 
         frame = Frame(image_path)
@@ -272,6 +288,9 @@ def prepare(self):
         if self.sample is not None and self.sample is not False:  # Nothing to do. Samples are ready
             return
 
+        if "test" in self.img_folder:
+            return  #The code for preparing test datasets exist but we are not doing that now
+
         if self.dataset_dir.endswith("_prepared") and not os.path.exists(self.dataset_dir.replace("_prepared", "")):
             return
 
@@ -294,7 +313,9 @@ def prepare(self):
 
 def main():
     """Main"""
-    crowd_human_dataset = CrowdHumanDataset(img_folder="CrowdHuman_train", ann_file="annotation_train.odgt")
+    crowd_human_dataset = CrowdHumanDataset(img_folder="CrowdHuman_test")
+    stuff = crowd_human_dataset[0]
+    stuff.get_view().render()
 
     crowd_human_dataset.prepare()
     for i, frames in enumerate(crowd_human_dataset.train_loader(batch_size=2, sampler=None, num_workers=0)):