Skip to content

Latest commit

 

History

History
111 lines (91 loc) · 4.5 KB

README.md

File metadata and controls

111 lines (91 loc) · 4.5 KB
Raw

EoID: A Pytorch Implementation

The official code for our paper End-to-End Zero-Shot HOI Detection via Vision and Language Knowledge Distillation (AAAI2023).

Installation

Install the dependencies. The code is tested with Pytorch 1.9.0.

bash prepare.sh

Data preparation

HICO-DET

HICO-DET dataset can be downloaded here. After finishing downloading, unpack the tarball (hico_20160224_det.tar.gz) to the data directory.

Instead of using the original annotations files, we use the annotation files provided by the PPDM authors. The annotation files can be downloaded from here. The downloaded annotation files have to be placed as follows.

data
 └─ hico_20160224_det
     |─ annotations
     |   |─ trainval_hico.json
     |   |─ test_hico.json
     |   └─ corre_hico.npy
     :

V-COCO

First clone the repository of V-COCO from here, and then follow the instruction to generate the file instances_vcoco_all_2014.json. Next, download the prior file prior.pickle from here. Place the files and make directories as follows.

CDN
 |─ data
 │   └─ v-coco
 |       |─ data
 |       |   |─ instances_vcoco_all_2014.json
 |       |   :
 |       |─ prior.pickle
 |       |─ images
 |       |   |─ train2014
 |       |   |   |─ COCO_train2014_000000000009.jpg
 |       |   |   :
 |       |   └─ val2014
 |       |       |─ COCO_val2014_000000000042.jpg
 |       |       :
 |       |─ annotations
 :       :

For our implementation, the annotation file have to be converted to the HOIA format. The conversion can be conducted as follows.

PYTHONPATH=data/v-coco \
        python convert_vcoco_annotations.py \
        --load_path data/v-coco/data \
        --prior_path data/v-coco/prior.pickle \
        --save_path data/v-coco/annotations

Note that only Python2 can be used for this conversion because vsrl_utils.py in the v-coco repository shows a error with Python3.

V-COCO annotations with the HOIA format, corre_vcoco.npy, test_vcoco.json, and trainval_vcoco.json will be generated to annotations directory.

Pre-trained model

Download the pretrained model of CLIP for CLIP50x16, and put it to the ckpt directory.

Download the pretrained model of DETR detector for ResNet50, and put it to the params directory.

python convert_parameters.py \
        --load_path params/detr-r50-e632da11.pth \
        --save_path params/detr-r50-pre-2stage-q64.pth \
        --num_queries 64

python convert_parameters.py \
        --load_path params/detr-r50-e632da11.pth \
        --save_path params/detr-r50-pre-2stage.pth \
        --dataset vcoco

Training

After the preparation, you can start training with the following commands. The trainings of EoID for HICO-DET under UA setting.

HICO-DET

bash train.sh

Evaluation

HICO-DET

You can conduct the evaluation with trained parameters for HICO-DET under UA setting as follows.

bash test.sh

Results

HICO-DET under UA setting

Full Seen Unseen Download
EoID(R50) 29.22 30.46 23.04 model

HICO-DET under UC setting

UC_Type Full Seen Unseen Download
EoID(R50) default 28.91 $\pm$ 0.33 30.39 $\pm$ 0.40 23.01 $\pm$ 1.98 model
EoID(R50) rare_first 29.52 31.39 22.04 model
EoID(R50)* rare_first 29.27 31.72 21.07 model
EoID(R50) non_rare_first 26.69 26.66 26.77 model
EoID(R50)* non_rare_first 25.64 25.35 26.80 model

* The original version of the model has been lost. This is a reimplement version on 2 RTX3090 with smaller batch size and learning rate.

Acknowledge

This repo is based on CDN, CLIP and ConsNet.