Skip to content

Latest commit

 

History

History
80 lines (63 loc) · 2.85 KB

EVALUATION.md

File metadata and controls

80 lines (63 loc) · 2.85 KB
Raw

Evaluation - Few-shot Semnatic Segmentation

Prepare FSS Benchmark

Download following datasets:

1. COCO-20i

Download COCO2014 train/val images and annotations:

wget http://images.cocodataset.org/zips/train2014.zip
wget http://images.cocodataset.org/zips/val2014.zip
wget http://images.cocodataset.org/annotations/annotations_trainval2014.zip

Download COCO2014 train/val annotations from this Google Drive: train2014.zip, val2014.zip. (and locate both train2014/ and val2014/ under annotations/ directory). Download data splits.

2. PASCAL-5i

Download PASCAL VOC2012 devkit (train/val data):

wget http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar

Download PASCAL VOC2012 SDS extended mask annotations from this Google Drive. Download data splits.

3. LVIS-92i

Download COCO2017 train/val images:

wget http://images.cocodataset.org/zips/train2017.zip
wget http://images.cocodataset.org/zips/val2017.zip

Download LVIS-92i extended mask annotations from our Google Drive: lvis.zip.

Create a directory 'datasets' for the above datasets and appropriately place each dataset to have following directory structure:

datasets/
├── fss/
    ├── COCO2014/ 
    │   ├── annotations/
    │   │   ├── train2014/
    │   │   └── val2014/
    │   ├── train2014/
    │   ├── val2014/
    │   └── splits
    │       ├── trn/
    │       └── val/
    ├── VOC2012/
    │   ├── Annotations/
    │   ├── ImageSets/
    │   ├── JPEGImages/
    │   ├── SegmentationClass/
    │   ├── SegmentationClassAug/
    │   ├── SegmentationObject/
    │   └── splits
    │       ├── trn/
    │       └── val/
    ├── LVIS/
    │   ├── coco/
    │   │   ├── train2017/
    │   │   └── val2017/
    │   ├── lvis_train.pkl
    │   └── lvis_val.pkl

Testing


CUDA_VISIBLE_DEVICES=0 python tools/eval_fss.py  \
  --benchmark coco \
  --fold 0 \
  --nshot 1 \
  --sine-weights /path/to/pt_sine_model/pytorch_model.bin \
  --log-root /path/to/fss/coco/1shot/fold0