Closed Morgan-117 closed 3 years ago
I also trained the network using LMDB format. You do not need the labelmap file. I do mean substraction (should be approximately (60,60,60)), and distortion (contrast , saturation, hue) as preprocessing.
Thanks for your help, I will try
Do you prepare the dataset like SSD(creat_data.sh, creat_annoset.py and convert_annoset)? In this way,I think we need a labelmap.prototxt file to convert the "112340"label to 4.
Please check bbox_util.cpp:
int state_digit_nr = 4; state = digits.at(state_digit_nr);
I extract the state from the complete classId
I want to prepared the DTLD dataset for training the model, I think I can use the class label(e.g. 112340) in the txt directly as you recommended. And then using lmdb format data created from lists of .jpg and .txt. So I don't need to using the labelmap file. By the way, do you use the same method for data augmentation as the original SSD? It will help a lot if you could share your data layer in your train.prototxt.