search

hizhangp / yolo_tensorflow

Tensorflow implementation of YOLO, including training and test phase.
MIT License
795 stars 442 forks source link

Did you train by yourself and test it? #62

Closed kashiwachen closed 3 years ago

kashiwachen commented 6 years ago

Hi, recently I train your net with my own dataset. However when I test with the training data, the result is nothing. I cannot understand it. So I want to know whether it works on your PC

bailiqun commented 6 years ago

met same problem with you..

NineBall9 commented 6 years ago

the same problem. There is no errors while training and the test picture output nothing.

lichacha commented 6 years ago

same

NineBall9 commented 6 years ago

Same! Does anyone can explain it.

haixiaotian commented 6 years ago

same

Tshzzz commented 6 years ago

maybe you didn't use pretrain file! see the config.py use weight

lichacha commented 6 years ago

I tested in VOC2007, testing yolosmall.ckpt directly is OK, training in voc2007 no output, and the use use pretrain file would be worse until there was no output. Are there too few training times？ I did 20k no output

bailiqun commented 6 years ago

have look with tensorflow/models/object_detection , may help

lichacha commented 6 years ago

Where is this file？ this project？

contactmat85 commented 6 years ago

same. I training with a new data set, with just one class. I use transfer learning initializing only the convolutional layer with yolosmall.ckpt. I have a good a quick convergency but no output, on the training set neither train

wheatjo commented 5 years ago

same,i have trained one class object, but output values are near zero

YunYang1994 commented 5 years ago

hope this helps you ! https://github.com/YunYang1994/tensorflow-yolov3

yanlongbinluck commented 5 years ago

same problem to you

Adienly commented 5 years ago

SAME, my loss is about 2(use my own dataset, imagesize is 90, cellsize is 2), when i test, there is nothing

kashiwachen commented 3 years ago

I already finished the project taking your framework as a reference. Happy develope!