gzz1529657064

> The first training part is for finetuning a model quickly. It freezes most layers, only to train on the last few layers. We can get an acceptable model for...

> 我现在测试ckpt和pb文件指标可以对齐了。加载模型的时候用不用ExponentialMovingAverage会有一点影响，在我这里是1个多点的影响。我这边主要是把image_demo.py里BGR转RGB那行去掉了，因为在utils.image_preprocess里做过了 确实如你所说，pb文件输出了正确的图片检测结果，但是图片的色彩发生了变换，和原图不一样，偏蓝，我觉得还需要转回来

> hi @gzz1529657064 do you have solved the problem? Yes, I use wrong prototxt, Now, I use models/yolov3/mobilenet_yolov3_train.prototxt and mobilenet_yolov3_test.prototxt. This prototxt data layer has only two top, data and...