• Python 2.7
• Pytorch 0.2 or higher
• CUDA 8.0
• Mask R-CNN: Follow the instructions of the mask-faster-rcnn repo, preparing everything needed for pyutils/mask-faster-rcnn.
• REFER API and data: Use the download links of REFER and go to the foloder running make. Follow data/README.md to prepare images and refcoco/refcoco+/refcocog annotations.
• COCO training set should be downloaded in pyutils/mask-faster-rcnn/data/coco/images/train2014.

python tools/prepro.py --dataset <DATASET> --splitBy <SPLITBY>

<DATASET> <SPLITBY> pairs contain: refcoco unc/refcoco+ unc/refcocog umd/refcocog google

1. Train the baseline segmentation model:

./experiments/scripts/train_baseline.sh <GPUID> <DATASET> <SPLITBY> <OUTPUT_POSTFIX>

Output model will be saved at <DATASET>_<SPLITBY>/output_<OUTPUT_POSTFIX>.

The Mask-RCNN model is in pyutils/mask-faster-rcnn/lib/nets/resnet_v1.py and pyutils/mask-faster-rcnn/lib/nets/network.py.

2. Train the model with spatial dynamic filters:

./experiments/scripts/train_spatial.sh <GPUID> <DATASET> <SPLITBY> <OUTPUT_POSTFIX>

The Mask-RCNN model is in pyutils/mask-faster-rcnn/lib/nets/resnet_v1_7f.py and pyutils/mask-faster-rcnn/lib/nets/network_7f.py.

3. Train the whole model with caption loss:

./experiments/scripts/train_cycle.sh <GPUID> <DATASET> <SPLITBY> att2in2 <CAPTION_LOSS_WEIGHT>

The whole model is in pyutils/mask-faster-rcnn/lib/nets/resnet_v1_cycle.py and pyutils/mask-faster-rcnn/lib/nets/network_cycle.py.

Losses are calculated in _add_losses().

1. Evaluate the baseline segmentation model:

./experiments/scripts/eval_baseline.sh <GPUID> <DATASET> <SPLITBY> <OUTPUT_POSTFIX> <MODEL_ITER>

Evaluate the model at <DATASET>_<SPLITBY>/output_<OUTPUT_POSTFIX>, of trained iteration <MODEL_ITER>.

2. Evaluate the model with spatial dynamic filters:

./experiments/scripts/eval_spatial.sh <GPUID> <DATASET> <SPLITBY> <OUTPUT_POSTFIX> <MODEL_ITER>