Skip to content

Commit

Permalink
update
Browse files Browse the repository at this point in the history
  • Loading branch information
@yjh0410
yjh0410 committed Feb 6, 2023
0 parents commit 4e9e13b
Show file tree
Hide file tree
Showing 85 changed files with 12,480 additions and 0 deletions.
8 changes: 8 additions & 0 deletions .gitignore
Original file line number Diff line number Diff line change
@@ -0,0 +1,8 @@
*.pt
*.pth
*.pkl
*.pyc
*.txt
__pycache__
det_results
.vscode
167 changes: 167 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,167 @@
# YOWOv2

## Requirements
- We recommend you to use Anaconda to create a conda environment:
```Shell
conda create -n yowo python=3.6
```

- Then, activate the environment:
```Shell
conda activate yowo
```

- Requirements:
```Shell
pip install -r requirements.txt
```

## Visualization

Comming soon ...

# Dataset
You can download **UCF24** from the following links:

## UCF101-24:
* Google drive

Link: https://drive.google.com/file/d/1Dwh90pRi7uGkH5qLRjQIFiEmMJrAog5J/view?usp=sharing

* BaiduYun Disk

Link: https://pan.baidu.com/s/11GZvbV0oAzBhNDVKXsVGKg

Password: hmu6

## AVA
You can use instructions from [here](https://github.com/yjh0410/AVA_Dataset) to prepare **AVA** dataset.

# Experiment
* UCF101-24

| Model | Clip | GFLOPs | Params | F-mAP | V-mAP | FPS | Weight |
|----------------|--------|--------|---------|-------|-------|---------|--------------|
| YOWOv2-Nano | 16 | 2.6 | 3.5 M | 78.8 | 48.0 | 42 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_nano_ucf24.pth) |
| YOWOv2-Tiny | 16 | 5.8 | 10.9 M | 80.5 | 51.3 | 50 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_tiny_ucf24.pth) |
| YOWOv2-Medium | 16 | 24.1 | 52.0 M | 83.1 | 50.7 | 42 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_medium_ucf24.pth) |
| YOWOv2-Large | 16 | 107.1 | 109.7 M | 85.2 | 52.0 | 30 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_large_ucf24.pth) |
| YOWOv2-Nano | 32 | 4.0 | 3.5 M | 79.4 | 49.0 | 42 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_nano_ucf24_k32.pth) |
| YOWOv2-Tiny | 32 | 9.0 | 10.9 M | 83.0 | 51.2 | 50 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_tiny_ucf24_k32.pth) |
| YOWOv2-Medium | 32 | 27.3 | 52.0 M | 83.7 | 52.5 | 40 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_medium_ucf24_k32.pth) |
| YOWOv2-Large | 32 | 183.9 | 109.7 M | 87.0 | 52.8 | 22 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_large_ucf24_k32.pth) |

* AVA v2.2

| Model | Clip | mAP | FPS | weight |
|----------------|------------|-----------|---------|--------------|
| YOWOv2-Nano | 16 | 12.6 | 40 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_nano_ava.pth) |
| YOWOv2-Tiny | 16 | 14.9 | 49 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_tiny_ava.pth) |
| YOWOv2-Medium | 16 | 18.4 | 41 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_medium_ava.pth) |
| YOWOv2-Large | 16 | 20.2 | 29 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_large_ava.pth) |
| YOWOv2-Nano | 32 | | | |
| YOWOv2-Tiny | 32 | 15.6 | 49 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_tiny_ava_k32.pth) |
| YOWOv2-Medium | 32 | 18.4 | 40 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_medium_ava_k32.pth) |
| YOWOv2-Large | 32 | 21.7 | 22 | [ckpt](https://github.com/yjh0410/YOWOv2/releases/download/yowo_v2_weight/yowo_v2_large_ava_k32.pth) |


## Train YOWOv2
* UCF101-24

```Shell
python train.py --cuda -d ucf24 --root path/to/dataset -v yowo_v2_nano --num_workers 4 --eval_epoch 1 --max_epoch 8 --lr_epoch 2 3 4 5 --lr 0.0001 -ldr 0.5 -bs 8 -accu 16
```

or you can just run the script:

```Shell
sh train_ucf.sh
```

* AVA
```Shell
python train.py --cuda -d ava_v2.2 --root path/to/dataset -v yowo_v2_nano --num_workers 4 --eval_epoch 1 --max_epoch 10 --lr_epoch 3 4 5 6 --lr 0.0001 -ldr 0.5 -bs 8 -accu 16 --eval
```

or you can just run the script:

```Shell
sh train_ava.sh
```

## Test YOWOv2
* UCF101-24
For example:

```Shell
python test.py --cuda -d ucf24 -v yowo_v2_nano --weight path/to/weight -size 224 --show
```

* AVA
For example:

```Shell
python test.py --cuda -d ava_v2.2 -v yowo_v2_nano --weight path/to/weight -size 224 --show
```

## Test YOWOv2 on AVA video
For example:

```Shell
python test_video_ava.py --cuda -d ava_v2.2 -v yowo_v2_nano --weight path/to/weight --video path/to/video --show
```

Note that you can set ```path/to/video``` to other videos in your local device, not AVA videos.

## Evaluate YOWOv2
* UCF101-24
For example:

```Shell
# Frame mAP
python eval.py \
--cuda \
-d ucf24 \
-v yowo_v2_nano \
-bs 16 \
-size 224 \
--weight path/to/weight \
--cal_frame_mAP \
```

```Shell
# Video mAP
python eval.py \
--cuda \
-d ucf24 \
-v yowo_v2_nano \
-bs 16 \
-size 224 \
--weight path/to/weight \
--cal_video_mAP \
```

* AVA

Run the following command to calculate frame mAP@0.5 IoU:

```Shell
python eval.py \
--cuda \
-d ava_v2.2 \
-v yowo_v2_nano \
-bs 16 \
--weight path/to/weight
```

## Demo
```Shell
# run demo
python demo.py --cuda -d ucf24 -v yowo_v2_nano -size 224 --weight path/to/weight --video path/to/video --show
-d ava_v2.2
```

## References
If you are using our code, please consider citing our paper.

Comming soon ...
21 changes: 21 additions & 0 deletions config/__init__.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
from .dataset_config import dataset_config
from .yowo_v2_config import yowo_v2_config


def build_model_config(args):
print('==============================')
print('Model Config: {} '.format(args.version.upper()))

if 'yowo_v2_' in args.version:
m_cfg = yowo_v2_config[args.version]

return m_cfg


def build_dataset_config(args):
print('==============================')
print('Dataset Config: {} '.format(args.dataset.upper()))

d_cfg = dataset_config[args.dataset]

return d_cfg
92 changes: 92 additions & 0 deletions config/dataset_config.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,92 @@
# Dataset configuration


dataset_config = {
'ucf24': {
# dataset
'gt_folder': './evaluator/groundtruths_ucf_jhmdb/groundtruths_ucf/',
# input size
'train_size': 224,
'test_size': 224,
# transform
'jitter': 0.2,
'hue': 0.1,
'saturation': 1.5,
'exposure': 1.5,
'sampling_rate': 1,
# cls label
'multi_hot': False, # one hot
# optimizer
'optimizer': 'adamw',
'momentum': 0.9,
'weight_decay': 5e-4,
# warmup strategy
'warmup': 'linear',
'warmup_factor': 0.00066667,
'wp_iter': 500,
# class names
'valid_num_classes': 24,
'label_map': (
'Basketball', 'BasketballDunk', 'Biking', 'CliffDiving',
'CricketBowling', 'Diving', 'Fencing', 'FloorGymnastics',
'GolfSwing', 'HorseRiding', 'IceDancing', 'LongJump',
'PoleVault', 'RopeClimbing', 'SalsaSpin', 'SkateBoarding',
'Skiing', 'Skijet', 'SoccerJuggling', 'Surfing',
'TennisSwing', 'TrampolineJumping', 'VolleyballSpiking', 'WalkingWithDog'
),
},

'ava_v2.2':{
# dataset
'frames_dir': 'frames/',
'frame_list': 'frame_lists/',
'annotation_dir': 'annotations/',
'train_gt_box_list': 'ava_v2.2/ava_train_v2.2.csv',
'val_gt_box_list': 'ava_v2.2/ava_val_v2.2.csv',
'train_exclusion_file': 'ava_v2.2/ava_train_excluded_timestamps_v2.2.csv',
'val_exclusion_file': 'ava_v2.2/ava_val_excluded_timestamps_v2.2.csv',
'labelmap_file': 'ava_v2.2/ava_action_list_v2.2_for_activitynet_2019.pbtxt', # 'ava_v2.2/ava_action_list_v2.2.pbtxt',
'class_ratio_file': 'config/ava_categories_ratio.json',
'backup_dir': 'results/',
# input size
'train_size': 224,
'test_size': 224,
# transform
'jitter': 0.2,
'hue': 0.1,
'saturation': 1.5,
'exposure': 1.5,
'sampling_rate': 1,
# cls label
'multi_hot': True, # multi hot
# train config
'optimizer': 'adamw',
'momentum': 0.9,
'weight_decay': 5e-4,
# warmup strategy
'warmup': 'linear',
'warmup_factor': 0.00066667,
'wp_iter': 500,
# class names
'valid_num_classes': 80,
'label_map': (
'bend/bow(at the waist)', 'crawl', 'crouch/kneel', 'dance', 'fall down', # 1-5
'get up', 'jump/leap', 'lie/sleep', 'martial art', 'run/jog', # 6-10
'sit', 'stand', 'swim', 'walk', 'answer phone', # 11-15
'brush teeth', 'carry/hold (an object)', 'catch (an object)', 'chop', 'climb (e.g. a mountain)', # 16-20
'clink glass', 'close (e.g., a door, a box)', 'cook', 'cut', 'dig', # 21-25
'dress/put on clothing', 'drink', 'drive (e.g., a car, a truck)', 'eat', 'enter', # 26-30
'exit', 'extract', 'fishing', 'hit (an object)', 'kick (an object)', # 31-35
'lift/pick up', 'listen (e.g., to music)', 'open (e.g., a window, a car door)', 'paint', 'play board game', # 36-40
'play musical instrument', 'play with pets', 'point to (an object)', 'press','pull (an object)', # 41-45
'push (an object)', 'put down', 'read', 'ride (e.g., a bike, a car, a horse)', 'row boat', # 46-50
'sail boat', 'shoot', 'shovel', 'smoke', 'stir', # 51-55
'take a photo', 'text on/look at a cellphone', 'throw', 'touch (an object)', 'turn (e.g., a screwdriver)', # 56-60
'watch (e.g., TV)', 'work on a computer', 'write', 'fight/hit (a person)', 'give/serve (an object) to (a person)', # 61-65
'grab (a person)', 'hand clap', 'hand shake', 'hand wave', 'hug (a person)', # 66-70
'kick (a person)', 'kiss (a person)', 'lift (a person)', 'listen to (a person)', 'play with kids', # 71-75
'push (another person)', 'sing to (e.g., self, a person, a group)', 'take (an object) from (a person)', # 76-78
'talk to (e.g., self, a person, a group)', 'watch (a person)' # 79-80
),
}
}
84 changes: 84 additions & 0 deletions config/yowo_v2_config.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,84 @@
# Model configuration


yowo_v2_config = {
'yowo_v2_nano': {
# backbone
## 2D
'backbone_2d': 'yolo_free_nano',
'pretrained_2d': True,
'stride': [8, 16, 32],
## 3D
'backbone_3d': 'shufflenetv2',
'model_size': '1.0x',
'pretrained_3d': True,
'memory_momentum': 0.9,
# head
'head_dim': 64,
'head_norm': 'BN',
'head_act': 'lrelu',
'num_cls_heads': 2,
'num_reg_heads': 2,
'head_depthwise': True,
},

'yowo_v2_tiny': {
# backbone
## 2D
'backbone_2d': 'yolo_free_tiny',
'pretrained_2d': True,
'stride': [8, 16, 32],
## 3D
'backbone_3d': 'shufflenetv2',
'model_size': '2.0x',
'pretrained_3d': True,
'memory_momentum': 0.9,
# head
'head_dim': 64,
'head_norm': 'BN',
'head_act': 'lrelu',
'num_cls_heads': 2,
'num_reg_heads': 2,
'head_depthwise': False,
},

'yowo_v2_medium': {
# backbone
## 2D
'backbone_2d': 'yolo_free_large',
'pretrained_2d': True,
'stride': [8, 16, 32],
## 3D
'backbone_3d': 'shufflenetv2',
'model_size': '2.0x',
'pretrained_3d': True,
'memory_momentum': 0.9,
# head
'head_dim': 128,
'head_norm': 'BN',
'head_act': 'silu',
'num_cls_heads': 2,
'num_reg_heads': 2,
'head_depthwise': False,
},

'yowo_v2_large': {
# backbone
## 2D
'backbone_2d': 'yolo_free_large',
'pretrained_2d': True,
'stride': [8, 16, 32],
## 3D
'backbone_3d': 'resnext101',
'pretrained_3d': True,
'memory_momentum': 0.9,
# head
'head_dim': 256,
'head_norm': 'BN',
'head_act': 'silu',
'num_cls_heads': 2,
'num_reg_heads': 2,
'head_depthwise': False,
},

}
Empty file added dataset/__init__.py
Empty file.
Loading

0 comments on commit 4e9e13b

Please sign in to comment.