Skip to content

Latest commit

 

History

History
72 lines (55 loc) · 2.44 KB

README.md

File metadata and controls

72 lines (55 loc) · 2.44 KB
Raw

Object Detection and Instance Segmentation

Detection and instance segmentation on MS COCO 2017 is implemented based on MMDetection. We follow settings and hyper-parameters of PVT and PoolFormer for the comparison,

Installation

Install mmcv-full and MMDetection v2.19.0, Later versions should work as well. The easiest way is to install via MIM

pip install -U openmim
mim install mmcv-full
mim install mmdet

Data preparation

Prepare COCO 2017 dataset according to the instructions in MMDetection. The dataset should be organized as

detection
├── data
│   ├── coco
│   │   ├── annotations
│   │   ├── train2017
│   │   ├── val2017
│   │   ├── test2017

ImageNet Pretraining

Put ImageNet-1K pretrained weights of backbone as

EfficientFormer
├── weights
│   ├── efficientformer_l1_300d.pth
│   ├── ...

Testing

Weights trained on COCO 2017 can be downloaded here. We provide a multi-GPU testing script, specify config file, checkpoint, and number of GPUs to use:

sh ./dist_test.sh config_file path/to/checkpoint #GPUs --eval bbox segm

For example, to test EfficientFormer-L1 on COCO 2017 on an 8-GPU machine,

sh ./dist_test.sh configs/mask_rcnn_efficientformer_l1_fpn_1x_coco.py path/to/efficientformer_l1_coco.pth 8 --eval bbox segm

Training

Single machine multi-GPU training

We provide PyTorch distributed data parallel (DDP) training script dist_train.sh, for example, to train EfficientFormer-L1 on an 8-GPU machine:

sh ./dist_train.sh configs/mask_rcnn_efficientformer_l1_fpn_1x_coco.py 8

Tips: specify configs and #GPUs!

Multi-node training

On Slurm-managed cluster, multi-node training can be launched by slurm_train.sh, similarly, to train EfficientFormer:

sh ./slurm_train.sh your-partition exp-name config-file work-dir

Tips: specify GPUs/CPUs/memory per node in the script slurm_train.sh based on your resource!