Gym pipeline

docs/changelog.md

@@ -6,17 +6,22 @@
 
 **New Features**
 - Automatically add modelzoo statistics to readthedocs ([#327](https://github.com/open-mmlab/mmaction2/pull/327))
+- Support GYM99 data preparation ([#331](https://github.com/open-mmlab/mmaction2/pull/331))
 
 **Improvements**
 - Support PyTorch 1.7 in CI ([#312](https://github.com/open-mmlab/mmaction2/pull/312))
 - Add random seed for building filelists ([#323](https://github.com/open-mmlab/mmaction2/pull/323))
 - Move docs about demo to `demo/README.md` ([#329](https://github.com/open-mmlab/mmaction2/pull/329))
+- Remove redundant code in `tools/test.py` ([#310](https://github.com/open-mmlab/mmaction2/pull/310))
 
 **Bug Fixes**
 - Fix a bug in BaseDataset when `data_prefix` is None ([#314](https://github.com/open-mmlab/mmaction2/pull/314))
 - Fix the bug of HVU object `num_classes` from 1679 to 1678 ([#307](https://github.com/open-mmlab/mmaction2/pull/307))
+- Fix OmniSource training configs ([#321](https://github.com/open-mmlab/mmaction2/pull/321))
+- Fix Issue #306: Bug of SampleAVAFrames ([#317](https://github.com/open-mmlab/mmaction2/pull/317))
 
 **ModelZoo**
+- Add SlowOnly model for GYM99, both RGB and Flow ([#336](https://github.com/open-mmlab/mmaction2/pull/336))
 
 ### v0.8.0 (31/10/2020)
 

tools/data/activitynet/download.py

@@ -2,13 +2,11 @@
 # https://github.com/activitynet/ActivityNet/blob/master/Crawler/Kinetics/download.py  # noqa: E501
 # The code is licensed under the MIT licence.
 import os
+import ssl
 import subprocess
 
 import mmcv
-
-import ssl  # isort:skip
-
-from joblib import Parallel, delayed  # isort:skip
+from joblib import Parallel, delayed
 
 ssl._create_default_https_context = ssl._create_unverified_context
 data_file = '../../../data/ActivityNet'

tools/data/gym/download.py

@@ -0,0 +1,99 @@
+# This scripts is copied from
+# https://github.com/activitynet/ActivityNet/blob/master/Crawler/Kinetics/download.py  # noqa: E501
+# The code is licensed under the MIT licence.
+import argparse
+import os
+import ssl
+import subprocess
+
+import mmcv
+from joblib import Parallel, delayed
+
+ssl._create_default_https_context = ssl._create_unverified_context
+
+
+def download(video_identifier,
+             output_filename,
+             num_attempts=5,
+             url_base='https://www.youtube.com/watch?v='):
+    """Download a video from youtube if exists and is not blocked.
+    arguments:
+    ---------
+    video_identifier: str
+        Unique YouTube video identifier (11 characters)
+    output_filename: str
+        File path where the video will be stored.
+    """
+    # Defensive argument checking.
+    assert isinstance(video_identifier, str), 'video_identifier must be string'
+    assert isinstance(output_filename, str), 'output_filename must be string'
+    assert len(video_identifier) == 11, 'video_identifier must have length 11'
+
+    status = False
+
+    if not os.path.exists(output_filename):
+        command = [
+            'youtube-dl', '--quiet', '--no-warnings', '--no-check-certificate',
+            '-f', 'mp4', '-o',
+            '"%s"' % output_filename,
+            '"%s"' % (url_base + video_identifier)
+        ]
+        command = ' '.join(command)
+        print(command)
+        attempts = 0
+        while True:
+            try:
+                subprocess.check_output(
+                    command, shell=True, stderr=subprocess.STDOUT)
+            except subprocess.CalledProcessError:
+                attempts += 1
+                if attempts == num_attempts:
+                    return status, 'Fail'
+            else:
+                break
+    # Check if the video was successfully saved.
+    status = os.path.exists(output_filename)
+    return status, 'Downloaded'
+
+
+def download_wrapper(youtube_id, output_dir):
+    """Wrapper for parallel processing purposes."""
+    # we do this to align with names in annotations
+    output_filename = os.path.join(output_dir, youtube_id + '.mp4')
+    if os.path.exists(output_filename):
+        status = tuple([youtube_id, True, 'Exists'])
+        return status
+
+    downloaded, log = download(youtube_id, output_filename)
+    status = tuple([youtube_id, downloaded, log])
+    return status
+
+
+def main(input, output_dir, num_jobs=24):
+    # Reading and parsing ActivityNet.
+    youtube_ids = mmcv.load(input).keys()
+    # Creates folders where videos will be saved later.
+    if not os.path.exists(output_dir):
+        os.makedirs(output_dir)
+    # Download all clips.
+    if num_jobs == 1:
+        status_list = []
+        for index in youtube_ids:
+            status_list.append(download_wrapper(index, output_dir))
+    else:
+        status_list = Parallel(n_jobs=num_jobs)(
+            delayed(download_wrapper)(index, output_dir)
+            for index in youtube_ids)
+
+    # Save download report.
+    mmcv.dump(status_list, 'download_report.json')
+
+
+if __name__ == '__main__':
+    description = 'Helper script for downloading GYM videos.'
+    p = argparse.ArgumentParser(description=description)
+    p.add_argument('input', type=str, help='The gym annotation file')
+    p.add_argument(
+        'output_dir', type=str, help='Output directory to save videos.')
+    p.add_argument('-n', '--num-jobs', type=int, default=24)
+    main(**vars(p.parse_args()))

tools/data/gym/download_annotations.sh

@@ -0,0 +1,14 @@
+#!/usr/bin/env bash
+
+set -e
+
+DATA_DIR="../../../data/gym/annotations"
+
+if [[ ! -d "${DATA_DIR}" ]]; then
+  echo "${DATA_DIR} does not exist. Creating";
+  mkdir -p ${DATA_DIR}
+fi
+
+wget https://sdolivia.github.io/FineGym/resources/dataset/finegym_annotation_info_v1.0.json -O $DATA_DIR/annotation.json
+wget https://sdolivia.github.io/FineGym/resources/dataset/gym99_train_element_v1.0.txt -O $DATA_DIR/gym99_train_org.txt
+wget https://sdolivia.github.io/FineGym/resources/dataset/gym99_val_element.txt -O $DATA_DIR/gym99_val_org.txt

tools/data/gym/download_videos.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+
+# set up environment
+conda env create -f environment.yml
+source activate gym
+pip install --upgrade youtube-dl
+
+DATA_DIR="../../../data/gym"
+ANNO_DIR="../../../data/gym/annotations"
+python download.py ${ANNO_DIR}/annotation.json ${DATA_DIR}/videos
+
+source deactivate gym
+conda remove -n gym --all

tools/data/gym/environment.yml

@@ -0,0 +1,36 @@
+name: kinetics
+channels:
+  - anaconda
+  - menpo
+  - conda-forge
+  - defaults
+dependencies:
+  - ca-certificates=2020.1.1
+  - certifi=2020.4.5.1
+  - ffmpeg=2.8.6
+  - libcxx=10.0.0
+  - libedit=3.1.20181209
+  - libffi=3.3
+  - ncurses=6.2
+  - openssl=1.1.1g
+  - pip=20.0.2
+  - python=3.7.7
+  - readline=8.0
+  - setuptools=46.4.0
+  - sqlite=3.31.1
+  - tk=8.6.8
+  - wheel=0.34.2
+  - xz=5.2.5
+  - zlib=1.2.11
+  - pip:
+    - decorator==4.4.2
+    - intel-openmp==2019.0
+    - joblib==0.15.1
+    - mkl==2019.0
+    - numpy==1.18.4
+    - olefile==0.46
+    - pandas==1.0.3
+    - python-dateutil==2.8.1
+    - pytz==2020.1
+    - six==1.14.0
+    - youtube-dl==2020.5.8

tools/data/gym/extract_frames.sh

@@ -0,0 +1,7 @@
+#!/usr/bin/env bash
+
+cd ../
+python build_rawframes.py ../../data/gym/subactions/ ../../data/gym/subaction_frames/ --level 1 --flow-type tvl1 --ext mp4 --task both --new-short 256
+echo "Raw frames (RGB and tv-l1) Generated"
+
+cd gym/

tools/data/gym/generate_file_list.py

@@ -0,0 +1,48 @@
+import os
+import os.path as osp
+
+annotation_root = '../../../data/gym/annotations'
+data_root = '../../../data/gym/subactions'
+frame_data_root = '../../../data/gym/subaction_frames'
+
+videos = os.listdir(data_root)
+videos = set(videos)
+
+train_file_org = osp.join(annotation_root, 'gym99_train_org.txt')
+val_file_org = osp.join(annotation_root, 'gym99_val_org.txt')
+train_file = osp.join(annotation_root, 'gym99_train.txt')
+val_file = osp.join(annotation_root, 'gym99_val.txt')
+train_frame_file = osp.join(annotation_root, 'gym99_train_frame.txt')
+val_frame_file = osp.join(annotation_root, 'gym99_val_frame.txt')
+
+train_org = open(train_file_org).readlines()
+train_org = [x.strip().split() for x in train_org]
+train = [x for x in train_org if x[0] + '.mp4' in videos]
+if osp.exists(frame_data_root):
+    train_frames = []
+    for line in train:
+        length = len(os.listdir(osp.join(frame_data_root, line[0])))
+        train_frames.append([line[0], str(length // 3), line[1]])
+    train_frames = [' '.join(x) for x in train_frames]
+    with open(train_frame_file, 'w') as fout:
+        fout.write('\n'.join(train_frames))
+
+train = [x[0] + '.mp4 ' + x[1] for x in train]
+with open(train_file, 'w') as fout:
+    fout.write('\n'.join(train))
+
+val_org = open(val_file_org).readlines()
+val_org = [x.strip().split() for x in val_org]
+val = [x for x in val_org if x[0] + '.mp4' in videos]
+if osp.exists(frame_data_root):
+    val_frames = []
+    for line in val:
+        length = len(os.listdir(osp.join(frame_data_root, line[0])))
+        val_frames.append([line[0], str(length // 3), line[1]])
+    val_frames = [' '.join(x) for x in val_frames]
+    with open(val_frame_file, 'w') as fout:
+        fout.write('\n'.join(val_frames))
+
+val = [x[0] + '.mp4 ' + x[1] for x in val]
+with open(val_file, 'w') as fout:
+    fout.write('\n'.join(val))

tools/data/gym/preparing_gym.md

@@ -0,0 +1,106 @@
+# Preparing GYM
+
+## Introduction
+
+```
+@inproceedings{shao2020finegym,
+  title={Finegym: A hierarchical video dataset for fine-grained action understanding},
+  author={Shao, Dian and Zhao, Yue and Dai, Bo and Lin, Dahua},
+  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  pages={2616--2625},
+  year={2020}
+}
+```
+
+For basic dataset information, please refer to the official [project](https://sdolivia.github.io/FineGym/) and the [paper](https://arxiv.org/abs/2004.06704).
+We currently provide the data pre-processing pipeline for GYM99.
+Before we start, please make sure that the directory is located at `$MMACTION2/tools/data/gym/`.
+
+## Step 1. Prepare Annotations
+
+First of all, you can run the following script to prepare annotations.
+
+```shell
+bash download_annotations.sh
+```
+
+## Step 2. Prepare Videos
+
+Then, you can run the following script to prepare videos.
+The codes are adapted from the [official crawler](https://github.com/activitynet/ActivityNet/tree/master/Crawler/Kinetics). Note that this might take a long time.
+
+```shell
+bash download_videos.sh
+```
+
+## Step 3. Trim Videos into Events.
+
+First, you need to trim long videos into events based on the annotation of GYM with the following scripts.
+
+```shell
+python trim_event.py
+```
+
+## Step 4. Trim Events into Subactions.
+
+Then, you need to trim events into subactions based on the annotation of GYM with the following scripts. We use the two stage trimming for better efficiency (trimming multiple short clips from a long video can be extremely inefficient, since you need to go over the video many times).
+
+```shell
+python trim_subaction.py
+```
+
+## Step 5. Extract RGB and Flow
+
+This part is **optional** if you only want to use the video loader for RGB model training.
+
+Before extracting, please refer to [install.md](/docs/install.md) for installing [denseflow](https://github.com/open-mmlab/denseflow).
+
+Run the following script to extract both rgb and flow using "tvl1" algorithm.
+
+```shell
+bash extract_frames.sh
+```
+
+## Step 6. Generate file list for GYM99 based on extracted subactions.
+
+You can use the following script to generate train / val lists for GYM99.
+
+```shell
+python generate_file_list.py
+```
+
+## Step 7. Folder Structure
+
+After the whole data pipeline for GYM preparation. You can get the subaction clips, event clips, raw videos and GYM99 train/val lists.
+
+In the context of the whole project (for GYM only), the full folder structure will look like:
+
+```
+mmaction2
+├── mmaction
+├── tools
+├── configs
+├── data
+│   ├── gym
+|   |   ├── annotations
+|   |   |   ├── gym99_train_org.txt
+|   |   |   ├── gym99_val_org.txt
+|   |   |   ├── gym99_train.txt
+|   |   |   ├── gym99_val.txt
+|   |   |   ├── annotation.json
+|   |   |   └── event_annotation.json
+│   │   ├── videos
+|   |   |   ├── 0LtLS9wROrk.mp4
+|   |   |   ├── ...
+|   |   |   └── zfqS-wCJSsw.mp4
+│   │   ├── events
+|   |   |   ├── 0LtLS9wROrk_E_002407_002435.mp4
+|   |   |   ├── ...
+|   |   |   └── zfqS-wCJSsw_E_006732_006824.mp4
+│   │   └── subactions
+|   |       ├── 0LtLS9wROrk_E_002407_002435_A_0003_0005.mp4
+|   |       ├── ...
+|   |       └── zfqS-wCJSsw_E_006244_006252_A_0000_0007.mp4
+```
+
+For training and evaluating on GYM, please refer to [getting_started](/docs/getting_started.md).