Thinklab-SJTU / R3Det_Tensorflow

Code for AAAI 2021 paper: R3Det: Refined Single-Stage Detector with Feature Refinement for Rotating Object
Apache License 2.0
546 stars 122 forks source link

关于论文中红绿灯检测的是如何训练的呢?我使用Retinanet作为 base model训练出来的效果并不理想 #145

Open kkkkf666 opened 2 years ago

yangxue0827 commented 2 years ago

用p2-p6得特征层试一下

kkkkf666 commented 2 years ago

@yangxue0827 不好意思我说错了,我现在是使用resnet50_coco_best_v2.1.0.h5作为base model,训练代码是kera-Retinanet那套进行训练,因为当时看readme中提供的是DOTA训练出来的R3Det*,预训练数据是遥感的并不是红绿灯的就没有使用,可以方便说一下要是用您这个训练红绿灯的应该怎么操作,是按照readme中操作进行吗?但是没有看到retinanet的Pretrain weights @yangxue0827 感谢您的回复


(1) Modify parameters (such as CLASS_NUM, DATASET_NAME, VERSION, etc.) in $PATH_ROOT/libs/configs/cfgs.py (2) Add category information in $PATH_ROOT/libs/label_name_dict/label_dict.py
(3) Add data_name to $PATH_ROOT/data/io/read_tfrecord_multi_gpu.py