Build your own target detection environment, model configuration, data configuration mmdetection

Data tagging

MMDetection

Project Download

The environment create

#  establish conda
conda create -n open-mmlab python=3.7 -y
conda activate open-mmlab

# pytorch  install  #  You can also migrate an environment directly conda create -n New --clone Old

conda install pytorch torchvision -c pytorch # conda install pytorch cudatoolkit=10.1 torchvision -c pytorch

#  Environmental installation 
pip install openmim
mim install mmdet

#  Error handling , See  https://github.com/open-mmlab/mmdetection/blob/master/docs/zh_cn/get_started.md/#Installation

A simple example

#%%

from mmdet.apis import init_detector, inference_detector, show_result_pyplot
import mmcv

#%%  Need from url Download pth file 

config_file = '../configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py'
# download the checkpoint from model zoo and put it in `checkpoints/`
# url: https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth
checkpoint_file = '../checkpoints/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth'

#%%

# build the model from a config file and a checkpoint file
model = init_detector(config_file, checkpoint_file, device='cuda:0')

#%%

# test a single image
img = 'demo.jpg'
result = inference_detector(model, img)

#%%

# show the results
show_result_pyplot(model, img, result)

Instructions for data preparation

COCO Format datasets

 {
            "id": 19, #  
            "width": 1280,
            "height": 720,
            "file_name": "013351.jpg", #  File name 
            "license": "",
            "flickr_url": "",
            "coco_url": "",
            "date_captured": ""
        },

The configuration file

• Super many configuration files ,config It contains all , When training, you only need to run train Sentence can be used

• open fast-rcnn A profile for
  

The model configuration

such as ：fastrcnn Refer to the ’…/base/models/fast_rcnn_r50_fpn.py’

…/base/models/fast_rcnn_r50_fpn.py The configuration is ：

# model settings
model = dict(  
    type='FastRCNN',
    backbone=dict(  #  Go to model In a folder backbone Folder 
        type='ResNet',# __init__ It is registered in the document type
        depth=50,#  Here are the parameters of the model 
        num_stages=4,
        out_indices=(0, 1, 2, 3),
        frozen_stages=1,
        norm_cfg=dict(type='BN', requires_grad=True),
        norm_eval=True,
        style='pytorch',
        init_cfg=dict(type='Pretrained', checkpoint='torchvision://resnet50')),
    neck=dict(
        type='FPN',
        in_channels=[256, 512, 1024, 2048],
        out_channels=256,
        num_outs=5),
    roi_head=dict(
        type='StandardRoIHead',
        bbox_roi_extractor=dict(
            type='SingleRoIExtractor',
            roi_layer=dict(type='RoIAlign', output_size=7, sampling_ratio=0),
            out_channels=256,
            featmap_strides=[4, 8, 16, 32]),
        bbox_head=dict(
            type='Shared2FCBBoxHead',
            in_channels=256,
            fc_out_channels=1024,
            roi_feat_size=7,
            num_classes=80,
            bbox_coder=dict(
                type='DeltaXYWHBBoxCoder',
                target_means=[0., 0., 0., 0.],
                target_stds=[0.1, 0.1, 0.2, 0.2]),
            reg_class_agnostic=False,
            loss_cls=dict(
                type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
            loss_bbox=dict(type='L1Loss', loss_weight=1.0))),
    # model training and testing settings
    train_cfg=dict(
        rcnn=dict(
            assigner=dict(
                type='MaxIoUAssigner',
                pos_iou_thr=0.5,
                neg_iou_thr=0.5,
                min_pos_iou=0.5,
                match_low_quality=False,
                ignore_iof_thr=-1),
            sampler=dict(
                type='RandomSampler',
                num=512,
                pos_fraction=0.25,
                neg_pos_ub=-1,
                add_gt_as_proposals=True),
            pos_weight=-1,
            debug=False)),
    test_cfg=dict(
        rcnn=dict(
            score_thr=0.05,
            nms=dict(type='nms', iou_threshold=0.5),
            max_per_img=100)))

Data configuration

data = dict(
    samples_per_gpu=2,#  Every gpu Upper batch_size  total batch= samples_per_gpu* numbergpu
    workers_per_gpu=2,#  One GPU Several processes load data 
    train=dict(
        proposal_file=data_root + 'proposals/rpn_r50_fpn_1x_train2017.pkl',
        pipeline=train_pipeline),
    val=dict(
        proposal_file=data_root + 'proposals/rpn_r50_fpn_1x_val2017.pkl',
        pipeline=test_pipeline),
    test=dict(
        proposal_file=data_root + 'proposals/rpn_r50_fpn_1x_val2017.pkl',
        pipeline=test_pipeline))

The most basic coco To configure

# dataset settings
dataset_type = 'CocoDataset'#  Which data set type 
data_root = 'data/coco/'#  Data path  
img_norm_cfg = dict(
    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
 
 # train Data processing of  
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations', with_bbox=True),
    dict(type='Resize', img_scale=(1333, 800), keep_ratio=True),#  keep aspect ratio  1333, 800 Is the size limit , No resize Target size to 
    dict(type='RandomFlip', flip_ratio=0.5),
    dict(type='Normalize', **img_norm_cfg),
    dict(type='Pad', size_divisor=32),
    dict(type='DefaultFormatBundle'),
    dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels']),#  Data information to be collected during training 
]
 # test Data processing of 
test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(
        type='MultiScaleFlipAug',
        img_scale=(1333, 800),
        flip=False,
        transforms=[
            dict(type='Resize', keep_ratio=True),
            dict(type='RandomFlip'),
            dict(type='Normalize', **img_norm_cfg),
            dict(type='Pad', size_divisor=32),
            dict(type='ImageToTensor', keys=['img']),
            dict(type='Collect', keys=['img']),
        ])
]
data = dict(
    samples_per_gpu=2,
    workers_per_gpu=2,
    train=dict(
        type=dataset_type,
        ann_file=data_root + 'annotations/instances_train2017.json',
        img_prefix=data_root + 'train2017/',
        pipeline=train_pipeline),
    val=dict(
        type=dataset_type, #  It's here dataset_type
        ann_file=data_root + 'annotations/instances_val2017.json',
        img_prefix=data_root + 'val2017/',
        pipeline=test_pipeline),
    test=dict(
        type=dataset_type,
        ann_file=data_root + 'annotations/instances_val2017.json',
        img_prefix=data_root + 'val2017/',
        pipeline=test_pipeline))
evaluation = dict(interval=1, metric='bbox')#  Interval between verifications , Every training one epoch Just verify it once  ,  The evaluation standard is  bbox

Training settings

 # model training and testing settings
    train_cfg=dict(
        rcnn=dict(
            assigner=dict(
                type='MaxIoUAssigner',
                pos_iou_thr=0.5,
                neg_iou_thr=0.5,
                min_pos_iou=0.5,
                match_low_quality=False,
                ignore_iof_thr=-1),
            sampler=dict(
                type='RandomSampler',
                num=512,
                pos_fraction=0.25,
                neg_pos_ub=-1,
                add_gt_as_proposals=True),
            pos_weight=-1,
            debug=False)),
    test_cfg=dict(
        rcnn=dict(
            score_thr=0.05,
            nms=dict(type='nms', iou_threshold=0.5),
            max_per_img=100)))

Reference and more

https://www.bilibili.com/video/BV1Hp4y1y788?

MMDetection Chinese document — Detailed explanation
MMDetection Chinese document —1. Ann loading
MMDetection Chinese document —2. introduction