Home

Awesome

Fusion Types and Cross Connection Study

Official Implementation used for training MATNet variants in our CVPR2022 Work. We provide the best models on MoCA with 70.6% mean IoU.

MoCA Results

Our paper uses threshold 0.2 and SOA MoCA comparison uses threshold 0.1. We recommend for better reporting on MoCA to compute area under the curve for different thresholds but is out of our current work scope.

MethodThFlipmIoUSR_0.5SR_0.6SR_0.7SR_0.8SR_0.9mSR
FusionSeg Modified0.2No42.347.943.635.924.29.439.2
RTNet0.2No60.767.962.453.643.423.950.2
MATNet reproduced0.2No67.375.970.861.948.626.056.6
MATNet NoBAR0.2No65.173.668.058.944.721.553.3
MATNet NoYTB0.2No54.759.953.544.031.013.440.3
MethodThFlipmIoUSR_0.5SR_0.6SR_0.7SR_0.8SR_0.9mSR
NonRecip CC + Gated Fusion0.1Yes70.279.474.164.649.023.858.2
NonRecip CC + Gated Fusion0.2Yes68.577.372.263.550.627.058.1
Recip CC + Gated Fusion0.2Yes70.681.275.565.048.123.058.6
Recip CC + Gated Fusion0.1Yes67.677.970.159.140.716.852.9
<div align="center"> <img src="https://github.com/MSiam/MATNet_FusionCrossConStudy/blob/main/figures/static_dynamic_recip.png" width="40%" height="40%"><br><br> </div> <div align="center"> <img src="https://github.com/MSiam/MATNet_FusionCrossConStudy/blob/main/figures/static_dynamic_masking_units.png" width="80%" height="40%"><br><br> </div>

Installation

The training and testing experiments are conducted using Python 3.7 PyTorch 1.9 with multi GPU support. Other minor Python modules can be installed by running

pip install -r requirements.txt

Datasets

Download Datasets

We follow MATNet and use the following two public available dataset for training. Here are some steps to prepare the data:

Prepare Edge Annotatios, HED Results and Optical Flow

Use MATNet instructions from here

Prepare MoCA for Evaluation

Follow motiongrouping instructions from here

Train

CUDA_VISIBLE_DEVICES=2,3 python train_MATNet.py -cfg_file CONFIG -gpu_id 0 1 -wandb_run_name WANDB_RUN

Test

python test_MATNet.py -ckpt_epoch BEST_EPOCH -ckpt_path CKPT_PATH -result_dir RESULT_DIR

Inference and Evaluation on MoCA

bash scripts/eval_MoCA.sh CFG CKPT BEST_EPOCH MASK_RESULT_DIR GPU_ID CSV_RESULT_DIR

Trained Models

For original MATNet use their provided models and for the reciprocal version with gated fusion that achieved best MoCA results use this model.

BibTeX

If you find this repository useful, please consider citing our work :t-rex:

  @InProceedings{kowal2022deeper,
   title={A Deeper Dive Into What Deep Spatiotemporal Networks Encode: Quantifying Static vs. Dynamic Information},
   author={Kowal, Matthew and Siam, Mennatullah and Islam, Md Amirul and Bruce, Neil and Wildes, Richard P. and Derpanis, Konstantinos G.},
   booktitle={Conference on Computer Vision and Pattern Recognition},
   year={2022}
 }
 

References