soccernet / sn-spotting Goto Github PK

Repository containing all necessary codes to get started on the SoccerNet Action Spotting challenge. This repository also contains several benchmark methods.

Python 99.73% Shell 0.27%

sn-spotting's Introduction

SoccerNet package

conda create -n SoccerNet python pip
conda activate SoccerNet
pip install SoccerNet
# pip install -e https://github.com/SoccerNet/SoccerNet
# pip install -e .

Structure of the data data for each game

SoccerNet main folder
- Leagues (england_epl/europe_uefa-champions-league/france_ligue-1/...)
  - Seasons (2014-2015/2015-2016/2016-2017)
    - Games (format: "{Date} - {Time} - {HomeTeam} {Score} {AwayTeam}")
      - SoccerNet-v2 - Labels / Manual Annotations
        
        video.ini: information on start/duration for each half of the game in the HQ video, in second
        
        Labels-v2.json: Labels from SoccerNet-v2 - action spotting
        
        Labels-cameras.json: Labels from SoccerNet-v1 - camera shot segmentation
      - SoccerNet-v2 - Videos / Automatically Extracted Features
        
        1_224p.mkv: 224p video 1st half - timmed with start/duration from HQ video - resolution 224*398 - 25 fps
        
        2_224p.mkv: 224p video 2nd half - timmed with start/duration from HQ video - resolution 224*398 - 25 fps
        
        1_720p.mkv: 720p video 1st half - timmed with start/duration from HQ video - resolution 720*1280 - 25 fps
        
        2_720p.mkv: 720p video 2nd half - timmed with start/duration from HQ video - resolution 720*1280 - 25 fps
        
        1_ResNET_TF2.npy: ResNET features @2fps for 1st half from SoccerNet-v2, extracted using TF2
        
        2_ResNET_TF2.npy: ResNET features @2fps for 2nd half from SoccerNet-v2, extracted using TF2
        
        1_ResNET_TF2_PCA512.npy: ResNET features @2fps for 1st half from SoccerNet-v2, extracted using TF2, with dimensionality reduced to 512 using PCA
        
        2_ResNET_TF2_PCA512.npy: ResNET features @2fps for 2nd half from SoccerNet-v2, extracted using TF2, with dimensionality reduced to 512 using PCA
        
        1_ResNET_5fps_TF2.npy: ResNET features @5fps for 1st half from SoccerNet-v2, extracted using TF2
        
        2_ResNET_5fps_TF2.npy: ResNET features @5fps for 2nd half from SoccerNet-v2, extracted using TF2
        
        1_ResNET_5fps_TF2_PCA512.npy: ResNET features @5fps for 1st half from SoccerNet-v2, extracted using TF2, with dimensionality reduced to 512 using PCA
        
        2_ResNET_5fps_TF2_PCA512.npy: ResNET features @5fps for 2nd half from SoccerNet-v2, extracted using TF2, with dimensionality reduced to 512 using PCA
        
        1_ResNET_25fps_TF2.npy: ResNET features @25fps for 1st half from SoccerNet-v2, extracted using TF2
        
        2_ResNET_25fps_TF2.npy: ResNET features @25fps for 2nd half from SoccerNet-v2, extracted using TF2
        
        1_player_boundingbox_maskrcnn.json: Player Bounding Boxes @2fps for 1st half, extracted with MaskRCNN
        
        2_player_boundingbox_maskrcnn.json: Player Bounding Boxes @2fps for 2nd half, extracted with MaskRCNN
        
        1_field_calib_ccbv.json: Field Camera Calibration @2fps for 1st half, extracted with CCBV
        
        2_field_calib_ccbv.json: Field Camera Calibration @2fps for 2nd half, extracted with CCBV
        
        1_baidu_soccer_embeddings.npy: Frame Embeddings for 1st half from https://github.com/baidu-research/vidpress-sports
        
        2_baidu_soccer_embeddings.npy: Frame Embeddings for 2nd half from https://github.com/baidu-research/vidpress-sports
      - Legacy from SoccerNet-v1
        
        Labels.json: Labels from SoccerNet-v1 - action spotting for goals/cards/subs only
        
        1_C3D.npy: C3D features @2fps for 1st half from SoccerNet-v1
        
        2_C3D.npy: C3D features @2fps for 2nd half from SoccerNet-v1
        
        1_C3D_PCA512.npy: C3D features @2fps for 1st half from SoccerNet-v1, with dimensionality reduced to 512 using PCA
        
        2_C3D_PCA512.npy: C3D features @2fps for 2nd half from SoccerNet-v1, with dimensionality reduced to 512 using PCA
        
        1_I3D.npy: I3D features @2fps for 1st half from SoccerNet-v1
        
        2_I3D.npy: I3D features @2fps for 2nd half from SoccerNet-v1
        
        1_I3D_PCA512.npy: I3D features @2fps for 1st half from SoccerNet-v1, with dimensionality reduced to 512 using PCA
        
        2_I3D_PCA512.npy: I3D features @2fps for 2nd half from SoccerNet-v1, with dimensionality reduced to 512 using PCA
        
        1_ResNET.npy: ResNET features @2fps for 1st half from SoccerNet-v1
        
        2_ResNET.npy: ResNET features @2fps for 2nd half from SoccerNet-v1
        
        1_ResNET_PCA512.npy: ResNET features @2fps for 1st half from SoccerNet-v1, with dimensionality reduced to 512 using PCA
        
        2_ResNET_PCA512.npy: ResNET features @2fps for 2nd half from SoccerNet-v1, with dimensionality reduced to 512 using PCA

How to Download Games (Python)

from SoccerNet.Downloader import SoccerNetDownloader

mySoccerNetDownloader = SoccerNetDownloader(LocalDirectory="path/to/soccernet")

# Download SoccerNet labels
mySoccerNetDownloader.downloadGames(files=["Labels.json"], split=["train", "valid", "test"]) # download labels
mySoccerNetDownloader.downloadGames(files=["Labels-v2.json"], split=["train", "valid", "test"]) # download labels SN v2
mySoccerNetDownloader.downloadGames(files=["Labels-cameras.json"], split=["train", "valid", "test"]) # download labels for camera shot

# Download SoccerNet features
mySoccerNetDownloader.downloadGames(files=["1_ResNET_TF2.npy", "2_ResNET_TF2.npy"], split=["train", "valid", "test"]) # download Features
mySoccerNetDownloader.downloadGames(files=["1_ResNET_TF2_PCA512.npy", "2_ResNET_TF2_PCA512.npy"], split=["train", "valid", "test"]) # download Features reduced with PCA
mySoccerNetDownloader.downloadGames(files=["1_player_boundingbox_maskrcnn.json", "2_player_boundingbox_maskrcnn.json"], split=["train", "valid", "test"]) # download Player Bounding Boxes inferred with MaskRCNN
mySoccerNetDownloader.downloadGames(files=["1_field_calib_ccbv.json", "2_field_calib_ccbv.json"], split=["train", "valid", "test"]) # download Field Calibration inferred with CCBV
mySoccerNetDownloader.downloadGames(files=["1_baidu_soccer_embeddings.npy", "2_baidu_soccer_embeddings.npy"], split=["train", "valid", "test"]) # download Frame Embeddings from https://github.com/baidu-research/vidpress-sports

# Download SoccerNet Challenge set (require password from NDA to download videos)
mySoccerNetDownloader.downloadGames(files=["1_ResNET_TF2.npy", "2_ResNET_TF2.npy"], split=["challenge"]) # download ResNET Features
mySoccerNetDownloader.downloadGames(files=["1_ResNET_TF2_PCA512.npy", "2_ResNET_TF2_PCA512.npy"], split=["challenge"]) # download ResNET Features reduced with PCA
mySoccerNetDownloader.downloadGames(files=["1_224p.mkv", "2_224p.mkv"], split=["challenge"]) # download 224p Videos (require password from NDA)
mySoccerNetDownloader.downloadGames(files=["1_720p.mkv", "2_720p.mkv"], split=["challenge"]) # download 720p Videos (require password from NDA)
mySoccerNetDownloader.downloadGames(files=["1_player_boundingbox_maskrcnn.json", "2_player_boundingbox_maskrcnn.json"], split=["challenge"]) # download Player Bounding Boxes inferred with MaskRCNN 
mySoccerNetDownloader.downloadGames(files=["1_field_calib_ccbv.json", "2_field_calib_ccbv.json"], split=["challenge"]) # download Field Calibration inferred with CCBV 
mySoccerNetDownloader.downloadGames(files=["1_baidu_soccer_embeddings.npy", "2_baidu_soccer_embeddings.npy"], split=["challenge"]) # download Frame Embeddings from https://github.com/baidu-research/vidpress-sports

# Download development kit per task
mySoccerNetDownloader.downloadDataTask(task="calibration-2023", split=["train", "valid", "test", "challenge"])
mySoccerNetDownloader.downloadDataTask(task="caption-2023", split=["train", "valid", "test", "challenge"])
mySoccerNetDownloader.downloadDataTask(task="jersey-2023", split=["train", "test", "challenge"])
mySoccerNetDownloader.downloadDataTask(task="reid-2023", split=["train", "valid", "test", "challenge"])
mySoccerNetDownloader.downloadDataTask(task="spotting-2023", split=["train", "valid", "test", "challenge"])
mySoccerNetDownloader.downloadDataTask(task="spotting-ball-2023", split=["train", "valid", "test", "challenge"], password=<PW_FROM_NDA>)
mySoccerNetDownloader.downloadDataTask(task="tracking-2023", split=["train", "test", "challenge"])

# Download SoccerNet videos (require password from NDA to download videos)
mySoccerNetDownloader.password = "Password for videos? (contact the author)"
mySoccerNetDownloader.downloadGames(files=["1_224p.mkv", "2_224p.mkv"], split=["train", "valid", "test"]) # download 224p Videos
mySoccerNetDownloader.downloadGames(files=["1_720p.mkv", "2_720p.mkv"], split=["train", "valid", "test"]) # download 720p Videos 
mySoccerNetDownloader.downloadRAWVideo(dataset="SoccerNet") # download 720p Videos 
mySoccerNetDownloader.downloadRAWVideo(dataset="SoccerNet-Tracking") # download single camera RAW Videos 

# Download SoccerNet in OSL ActionSpotting format
mySoccerNetDownloader.downloadDataTask(task="spotting-OSL", split=["train", "valid", "test", "challenge"], version="ResNET_PCA512")
mySoccerNetDownloader.downloadDataTask(task="spotting-OSL", split=["train", "valid", "test", "challenge"], version="baidu_soccer_embeddings")
mySoccerNetDownloader.downloadDataTask(task="spotting-OSL", split=["train", "valid", "test", "challenge"], version="224p", password=<PW_FROM_NDA>)

How to read the list Games (Python)

from SoccerNet.utils import getListGames
print(getListGames(split="train")) # return list of games recommended for training
print(getListGames(split="valid")) # return list of games recommended for validation
print(getListGames(split="test")) # return list of games recommended for testing
print(getListGames(split="challenge")) # return list of games recommended for challenge
print(getListGames(split=["train", "valid", "test", "challenge"])) # return list of games for training, validation and testing
print(getListGames(split="v1")) # return list of games from SoccerNetv1 (train/valid/test)

sn-spotting's People

Contributors

Stargazers

Watchers

Forkers

ddk1246 huichuanliu phuc16102001 changewow hmkhalla bhosalems minghua3 sangbumchoi adjuan

sn-spotting's Issues

Meaning of "confidence" field in prediction file for evaluation

Hi,

It is quite an interesting soccer challenge. Though I would like to understand one small point about "confidence" field in prediction files.

During evaluation, there is a necessity to provide a confidence of the predicted event. I was wondering how this value affect the evaluation results - Average-mAP (e.g. whether the predicted events are thresholded in "evaluate" method or something else)?

Thank you.

Cannot download data for spotting-ball-2024 (HTTP Error 401: Unauthorized)

As suggested in the title, I am unable to download the action spotting dataset for the 2024 challenge. I get a 'HTTP Error 401: Unauthorized'. I tried using the 'download_ball_data.py' from the official repo, inputting the password provided after signing the NDA, using both python 3.9 and 3.10.

Can you please advise what to do here?

Using a different version of libraries

Running `pip3 install torch==1.6.0 torchvision==0.7.0 -f https://download.pytorch.org/whl/cu101/torch_stable.html`

Gives.................

Looking in links: https://download.pytorch.org/whl/cu101/torch_stable.html
ERROR: Could not find a version that satisfies the requirement torch==1.6.0 (from versions: 1.8.0, 1.8.1, 1.9.0, 1.10.0, 1.10.1, 1.10.2, 1.11.0, 1.12.0, 1.12.1, 1.13.0, 1.13.1, 2.0.0)
ERROR: No matching distribution found for torch==1.6.0

Running `conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.1 -c pytorch`

Gives....................

Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.

PackagesNotFoundError: The following packages are not available from current channels:

cudatoolkit=10.1

pytorch==1.6.0

torchvision==0.7.0

Current channels:

https://conda.anaconda.org/pytorch/linux-aarch64

https://conda.anaconda.org/pytorch/noarch

https://conda.anaconda.org/conda-forge/linux-aarch64

https://conda.anaconda.org/conda-forge/noarch

https://repo.anaconda.com/pkgs/main/linux-aarch64

https://repo.anaconda.com/pkgs/main/noarch

https://repo.anaconda.com/pkgs/r/linux-aarch64

https://repo.anaconda.com/pkgs/r/noarch

To search for alternate channels that may provide the conda package you're
looking for, navigate to
https://anaconda.org
and use the search bar at the top of the page.

Since I cannot install the required versions, will installing higher versions break the code?

Thanks for your amazing work.

CALF model training error !

@SilvioGiancola I want to ask that the CALF model training is throwing error with


RuntimeError: CUDA out of memory. Tried to allocate 7.38 GiB (GPU 0; 10.76 GiB total capacity; 7.45 GiB already allocated; 2.19 GiB free; 7.45 GiB reserved in total by PyTorch)
Exception in thread Thread-2:

while there is no any of error with Temporary pooling model NetVLAD++

Can you or any friend guide me why is it so. I tried to free cache but didn't work. I used smaller batch_size and also failed to resolve so. If you can kindly help.

Baidu feats extractor

Is there any feature extractor routine available for baidu embeddings ? How these features are extracted?

ABOUT CVPR22 SUBMISSION

Hi, i am wondering when will the cvpr22 evaluation server be ready.

I have just participated in this challenge and I find that there is no submission page on EvalAI.

By the way, are challenge set games included in Baidu features? Thx.

Possible OOM issue in training for NetVLAD++ (Linux Ubuntu 18.04)

Hello!
Thank you so much for providing this open repository for everyone!

I would like to ask about my issue running training for the Benchmark model.
I tried training for the NetVLAD++ benchmark model in my system and ran into this issue:

This is the parameters for the training:

2023-08-04 11:16:14,511 [MainThread ] [INFO ] Parameters:
2023-08-04 11:16:14,511 [MainThread ] [INFO ] SoccerNet_path : /media/ogatalab/SSD-PGU3C/SoccerNet
2023-08-04 11:16:14,511 [MainThread ] [INFO ] features : ResNET_TF2.npy
2023-08-04 11:16:14,511 [MainThread ] [INFO ] max_epochs : 1000
2023-08-04 11:16:14,511 [MainThread ] [INFO ] load_weights : None
2023-08-04 11:16:14,511 [MainThread ] [INFO ] model_name : NetVLAD++
2023-08-04 11:16:14,511 [MainThread ] [INFO ] test_only : False
2023-08-04 11:16:14,511 [MainThread ] [INFO ] split_train : ['train']
2023-08-04 11:16:14,511 [MainThread ] [INFO ] split_valid : ['valid']
2023-08-04 11:16:14,511 [MainThread ] [INFO ] split_test : ['test', 'challenge']
2023-08-04 11:16:14,511 [MainThread ] [INFO ] version : 2
2023-08-04 11:16:14,511 [MainThread ] [INFO ] feature_dim : None
2023-08-04 11:16:14,511 [MainThread ] [INFO ] evaluation_frequency : 10
2023-08-04 11:16:14,511 [MainThread ] [INFO ] framerate : 2
2023-08-04 11:16:14,511 [MainThread ] [INFO ] window_size : 15
2023-08-04 11:16:14,511 [MainThread ] [INFO ] pool : NetVLAD++
2023-08-04 11:16:14,511 [MainThread ] [INFO ] vocab_size : 64
2023-08-04 11:16:14,511 [MainThread ] [INFO ] NMS_window : 30
2023-08-04 11:16:14,511 [MainThread ] [INFO ] NMS_threshold : 0.0
2023-08-04 11:16:14,511 [MainThread ] [INFO ] batch_size : 256
2023-08-04 11:16:14,511 [MainThread ] [INFO ] LR : 0.001
2023-08-04 11:16:14,511 [MainThread ] [INFO ] LRe : 1e-06
2023-08-04 11:16:14,511 [MainThread ] [INFO ] patience : 10
2023-08-04 11:16:14,511 [MainThread ] [INFO ] GPU : 0
2023-08-04 11:16:14,511 [MainThread ] [INFO ] max_num_worker : 4
2023-08-04 11:16:14,511 [MainThread ] [INFO ] seed : 0
2023-08-04 11:16:14,511 [MainThread ] [INFO ] loglevel : INFO

I have tried to reduce the batch_size and max_num_workers as well, but the problem persists.
Also, not sure if it is needed, but it seems it only happens after it runs the test() procedure.

I found no trouble training with PCA512 features, so I figure there must be some resource issues.
When I look at the System Monitor and GPU usage, this is the reading of the result.

I have tried increasing the swap file to more than 64GB, but it only results in my system freezing and crashing.
Are you familiar with this issue? Is there any way for me to bypass this?
Or, if it is no trouble, can you tell me the recommended specification to run this training?
Thank you!

My system specs (RAM 16GB, SSD 2TB):

Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
CPU(s): 6
On-line CPU(s) list: 0-5
Thread(s) per core: 1
Core(s) per socket: 6
Socket(s): 1
NUMA node(s): 1
Vendor ID: AuthenticAMD
CPU family: 23
Model: 113
Model name: AMD Ryzen 5 3500 6-Core Processor
Stepping: 0
CPU MHz: 3266.777
CPU max MHz: 3600.0000
CPU min MHz: 2200.0000
BogoMIPS: 7186.49
Virtualization: AMD-V
L1d cache: 32K
L1i cache: 32K
L2 cache: 512K
L3 cache: 8192K
NUMA node0 CPU(s): 0-5

GPU (NVIDIA GeForce RTX 3060 Ti)

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.182.03 Driver Version: 470.182.03 CUDA Version: 11.4 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce ... Off | 00000000:06:00.0 On | N/A |
| 30% 51C P2 57W / 200W | 2270MiB / 7973MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 1067 G /usr/bin/gnome-shell 87MiB |
| 0 N/A N/A 1431 G /usr/lib/xorg/Xorg 146MiB |
| 0 N/A N/A 1546 G /usr/bin/gnome-shell 46MiB |
| 0 N/A N/A 2130 G ...RendererForSitePerProcess 87MiB |
| 0 N/A N/A 2754 G /usr/bin/nvidia-settings 0MiB |
| 0 N/A N/A 5116 C python 1809MiB |
| 0 N/A N/A 23313 G ...464382835143902905,262144 77MiB |
+-----------------------------------------------------------------------------+

Evaluation Code

Hi,

Thanks for this awesome dataset!

I would like to go through the evaluation code. However, I am unable to find it. Can you please point me out the location where I can find the code to calculate Average-mAP?

Wrong Interpretation

I have to ask one question. I used NetVlad++ model and TempPooling inference results on my own data video test is that accuracy is less than CALF model json. All the events are less than value 0,25 in visibility. Could you give any hint why it could be. I guess that I need correct K_V2 values

Test on external video

I clone repo and I run

user@PC MINGW64 ~
$ cd GitProjects/sn-spotting/Benchmarks/CALF/

user@PC MINGW64 ~/GitProjects/sn-spotting/Benchmarks/CALF (main)
$ ls
AUTHORS  LICENSE  README.md  img  inference  models  outputs  src

user@PC MINGW64 ~/GitProjects/sn-spotting/Benchmarks/CALF (main)
$ python inference/main.py --video_path="test.mp4" --model_name CALF_benchmark
Traceback (most recent call last):
  File "~\GitProjects\sn-spotting\Benchmarks\CALF\inference\main.py", line 10, in <module>
    from dataset import SoccerNetClipsTesting
  File "~\GitProjects\sn-spotting\Benchmarks\CALF\inference\dataset.py", line 21, in <module>
    from Features.VideoFeatureExtractor import VideoFeatureExtractor, PCAReducer
ModuleNotFoundError: No module named 'Features'

Error while training!

I have just tried to reproduce the code and I faced this issue. Can you guess what could be the reason? Is there something in download of the dataset? Or?

Train 79: Time 0.043s (it:0.011s) Data:0.027s (it:0.000s) Loss 7.8882e-02 : 100%|███████████████████████████████████████| 431/431 [00:18<00:00, 23.24it/s]
Evaluate 79: Time 0.042s (it:0.005s) Data:0.033s (it:0.001s) Loss 1.0849e-01 : 100%|████████████████████████████████████| 144/144 [00:05<00:00, 24.48it/s]
2022-04-01 12:20:21,465 [MainThread ] [INFO ] Plateau Reached and no more reduction -> Exiting Loop
2022-04-01 12:20:21,487 [MainThread ] [INFO ] Checking/Download features and labels locally
Test (spot.): Time 2.487s (it:2.519s) Data:2.111s (it:2.149s) : 67%|██████████████████████████████████▊ | 67/100 [02:46<01:22, 2.49s/it]
Traceback (most recent call last):
File "/home/main/Desktop/soccer/SoccerNetv2-DevKit/Task1-ActionSpotting/TemporallyAwarePooling/src/main.py", line 174, in
main(args)
File "/home/main/Desktop/soccer/SoccerNetv2-DevKit/Task1-ActionSpotting/TemporallyAwarePooling/src/main.py", line 86, in main
results = testSpotting(test_loader, model=model, model_name=args.model_name, NMS_window=args.NMS_window, NMS_threshold=args.NMS_threshold)
File "/home/main/Desktop/soccer/SoccerNetv2-DevKit/Task1-ActionSpotting/TemporallyAwarePooling/src/train.py", line 218, in testSpotting
for i, (game_ID, feat_half1, feat_half2, label_half1, label_half2) in t:
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/tqdm/std.py", line 1180, in iter
for obj in iterable:
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 521, in next
data = self._next_data()
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 1203, in _next_data
return self._process_data(data)
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 1229, in _process_data
data.reraise()
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/torch/_utils.py", line 434, in reraise
raise exception
ValueError: Caught ValueError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/torch/utils/data/_utils/worker.py", line 287, in _worker_loop
data = fetcher.fetch(index)
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/torch/utils/data/_utils/fetch.py", line 49, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/torch/utils/data/utils/fetch.py", line 49, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/home/main/Desktop/soccer/SoccerNetv2-DevKit/Task1-ActionSpotting/TemporallyAwarePooling/src/dataset.py", line 190, in getitem
feat_half1 = np.load(os.path.join(self.path, self.listGames[index], "1" + self.features))
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/numpy/lib/npyio.py", line 430, in load
return format.read_array(fid, allow_pickle=allow_pickle,
File "/home/main/anaconda3/envs/soccer/lib/python3.9/site-packages/numpy/lib/format.py", line 786, in read_array
array.shape = shape
ValueError: cannot reshape array of size 989152 into shape (5400,2048)

CALF_Calibration_GCN: Code & RuntimeError

I am trying to run CALF_Calibration_GCN https://github.com/SoccerNet/sn-spotting/tree/main/Benchmarks/CALF_Calibration_GCN to replicate the Benchmark results.
In your code to run this model, model_name is specified twice. I presume the second, --model_name=calib_GCN_run_${i}, is redundant here?:

python src/main.py \
--SoccerNet_path=path/to/SoccerNet/ \
--features=ResNET_TF2_PCA512.npy \
--num_features=512 \
--model_name=calib_GCN \
--batch_size 32 \
--evaluation_frequency 20 \
--chunks_per_epoch 18000 \
--model_name=calib_GCN_run_${i}  \
--backbone_feature=2DConv \
--backbone_player=resGCN-14 \
--dist_graph_player=25 \
--calibration \
--feature_multiplier 2 \
--class_split visual

Also, even if this line is removed, I keep getting RuntimeError: CUDA out of memory.
I have tried reducing batch_size from 32 to 16 and running the model on a server node with a greater number of GPUs, but without success.
Do you have any suggestions as to how I can prevent this error?

DISCORD INVITE INVALID

Hi! I fail to join the SoccerNet discord and it seems that the invitation URL is expired. Could you send me a new invite link? I wonder the final results on the challenge set. Thank you!

CALF Calibration & CALF Calibration GCN Benchmark code error

While trying to run the benchmark CALF_Calibration_GCN code as per the instructions in this repository, I receive the following error:

Traceback (most recent call last):
  File "src/main.py", line 10, in <module>
    from dataset import SoccerNetClips, SoccerNetClipsTesting, collateGCN, collateGCNTesting
  File "/work2/b_rory/work/SoccerNet/sn-spotting/Benchmarks/CALF_Calibration_GCN/src/dataset.py", line 24, in <module>
    from torch_geometric.data import Data
  File "/home/b_rory/workspace2/virtualenvs/venv/envs/CALF-pytorch/lib/python3.8/site-packages/torch_geometric/__init__.py", line 4, in <module>
    import torch_geometric.data
  File "/home/b_rory/workspace2/virtualenvs/venv/envs/CALF-pytorch/lib/python3.8/site-packages/torch_geometric/data/__init__.py", line 7, in <module>
    from .lightning_datamodule import (
  File "/home/b_rory/workspace2/virtualenvs/venv/envs/CALF-pytorch/lib/python3.8/site-packages/torch_geometric/data/lightning_datamodule.py", line 8, in <module>
    from torch_geometric.loader.dataloader import DataLoader
  File "/home/b_rory/workspace2/virtualenvs/venv/envs/CALF-pytorch/lib/python3.8/site-packages/torch_geometric/loader/__init__.py", line 15, in <module>
    from .dynamic_batch_sampler import DynamicBatchSampler
  File "/home/b_rory/workspace2/virtualenvs/venv/envs/CALF-pytorch/lib/python3.8/site-packages/torch_geometric/loader/dynamic_batch_sampler.py", line 9, in <module>
    class DynamicBatchSampler(torch.utils.data.sampler.Sampler[List[int]]):
TypeError: 'type' object is not subscriptable

Could you please advise how I can resolve this?

Multiple GPU usage

Is it possible to use the multi GPU with the routine available? with shared memory ?

Features extraction

I have to ask one question. How to extract from any soccer match the metadata in json with the features npy file with the time and its spotting feature info? like label_camera json files are from any of the trained engine CALF? The routine is available in the list_management.py in utilities. Is this possible while extracting features?