xiaochus / yolov3 Goto Github PK

I have noticed that Opencv imread was loading color channels in BGR order, what I did not know before. Since I suppose the original weights are made to deal with the channels in RGB order, I swapped the channels back by adding the line:

image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

,at the beggining of the function process_image.

It indeed increased the detection rate especially for the yolo_tiny model. It may solve issue #8 btw.

When i convert my own trained yolov3.weights to .h5, the error will shown as follows:
Traceback (most recent call last):
File "yad2k.py", line 275, in
_main(parser.parse_args())
File "yad2k.py", line 151, in _main
buffer=weights_file.read(weights_size * 4))
TypeError: buffer is too small for requested array

How to solve this problem
Thanks!

Hi,

I was looking at the original YOLO cfg here. It seems the original one accepts a input of size 608x608, not 416 as mentioned in your config file.

So, how is this even working considering you are downloading the weights for for the larger 608x608 image. I understand that the initial convolutions will be the same but the size of your outputs will be different. Am I mistaking something here?

Update: looks like the cfg file was changed 4 days ago. But still the question remains as to why it worked given the weights would have been from the newer model?

你好，我使用您的模型对一个文件夹下数万张图片进行打标，发现每个for循环的时间比前一次时间会增加，最开始是6s一张，机器跑一个通宵后变成100s一张，我停止程序后立即启动又变成6s，然后一次增加时长。检测程序发现是yolo_model.py中
耗时较长，请问应该如何解决呢？

I have a better code about yolo https://github.com/Stinky-Tofu/Stronger-yolo

I found mistake in the code.

File: yolo_model.py

for image with width not equal to height, this lines causes to error (lines 59-63):

        col = np.tile(np.arange(0, grid_w), grid_w).reshape(-1, grid_w)
        row = np.tile(np.arange(0, grid_h).reshape(-1, 1), grid_h)

        col = col.reshape(grid_h, grid_w, 1, 1).repeat(3, axis=-2)
        row = row.reshape(grid_h, grid_w, 1, 1).repeat(3, axis=-2)

instead, copy and paste these lines :

col = np.tile(np.arange(0, grid_w), grid_h).reshape(-1, grid_w)
row = np.tile(np.arange(0, grid_h).reshape(-1, 1), grid_w)
col = col.reshape(grid_h, grid_w, 1, 1).repeat(3, axis=-2)
row = row.reshape(grid_h, grid_w, 1, 1).repeat(3, axis=-2)

@xiaochus hi,xiaochus.when i tested the video with my own trained model,i couldn't identify anything;instead,i could identify if i used the author's yolov3.weights.i dont know why this result was produced,cause when i use my trained model on darknet,it worked well.

这个代码可以用usb摄像头实时检测吗

I found that you use:
col = np.tile(np.arange(1, grid_w + 1), grid_w).reshape(-1, grid_w)
row = np.tile(np.arange(1, grid_h + 1).reshape(-1, 1), grid_h)

It means you use offsets of [1, W] and [1, H], But why you not use [0, W-1] and [0, H-1]?

How could this be possible?
I am using the demo.py script.

Good work!
But I think there is a bug in darknet53.py, line 51:
in stead of x = conv2d_unit(inputs, 2 * filters, (3, 3)), it shoud be x = conv2d_unit(x, 2 * filters, (3, 3)).
Can you check it?

how to solve this ??????

ValueError: Error when checking : expected input_1 to have shape (608, 608, 3) but got array with shape (416, 416, 3)

I am using this repo to run YoloV3 on a test video. My GPU support is GeForce GTX 1080 Ti. The video is playing with almost 3 sec per frame. I have no idea why it is so slow, even though my GPU is good enough. I have attached a screen-shot of my terminal while running the test video. Any idea or guidance is much appreciated. Thank You

@xiaochus

这个repo只有detection功能，那么怎么实现这个功能的呢？
写的一堆yolo model代码一点没用到，他只是用别人写好的yad2k.py文件把darknet.weight文件转成keras的.h5文件，然后load_model，就开始其实就是用训练好的模型，predict然后完事了。
所以我觉得没必要再看了，没有train功能，前面star的不知道真的看没看。

I found an error in this code. I am looking forward to the author's reply thanks
File "demo.py", line 152, in
image = detect_image(image, yolo, all_classes)
File "demo.py", line 90, in detect_image
boxes, classes, scores = yolo.predict(pimage, image.shape)
File "C:\Users\Administrator\Desktop\YOLOv3-master\model\yolo_model.py", line 209, in predict
boxes, classes, scores = self._yolo_out(outs, shape)
File "C:\Users\Administrator\Desktop\YOLOv3-master\model\yolo_model.py", line 158, in _yolo_out
b, c, s = self._process_feats(out, anchors, mask)
File "C:\Users\Administrator\Desktop\YOLOv3-master\model\yolo_model.py", line 54, in _process_feats
box_wh = box_wh * anchors_tensor
ValueError: operands could not be broadcast together with shapes (13,13,2) (1,1,3,2)

I've trained yolov3 on mobilenet backend and get a good MAP(about 90M and 2 times faster than keras yolov3).I upload my code for mobilenet-yolov3 on https://github.com/Adamdad/keras-YOLOv3-mobilenet.git
use it as you wish

Hi,I am trying to convert Yolo wieghts to Keras on my Mac but i can't run the script. It says;

Traceback (most recent call last):
File "yad2k.py", line 9, in
import configparser
ImportError: No module named configparser

i tried pip install configparser but it didn't work. Do you have any advise?

广告虽可耻，大家可以看看我复现的这版yolov4，精心设计了损失函数，训练了70000步就没出现过nan，mAP至少可达36%，欢迎点star：

https://github.com/miemie2013/Keras-YOLOv4

Hello,

I'm trying to train a network for different tasks, based on your project. The main issue I have is that if I create the same net with random weights instead of those from yolov3.weights (using this script: https://pastebin.com/RECciwrs) I get a net with the same number of params, the same layers, but on inference it outputs a matrix of NaNs. Since it outputs NaNs, I cannot train it, it will not converge to the desired weights.

How to repro:

1. Save the script from https://pastebin.com/RECciwrs as create_yolo_from_scratch.py
2. python yad2k.py cfg\yolo.cfg yolov3.weights data\yolo.h5
3. python create_yolo_from_scratch.py cfg/yolo.cfg yolo_random_weights.h5
4. For convenience, resize giraffe.jpg to 416x416 and rename as giraffe416.jpg
5. Run in terminal:
   python

from keras.models import load_model
import numpy as np
import cv2
yolo = load_model('yolo.h5')
yolo_rw = load_model('yolo_random_weights.h5')
path = 'images/test/giraffe416.jpg'
image = cv2.imread(path)
image = np.array(image, dtype='float32')
image /= 255.
image = np.expand_dims(image, axis=0)
yolo.predict(image)
# will output a matrix of numbers
yolo_rw.predict(image)
# will output a matrix of NaNs
yolo_rw.set_weights(yolo.get_weights())
yolo_rw.predict(image)
# will output a matrix of numbers

Without using yolov3.weights, do you have any advice on how to seed the network so that it's trainable? I want to then use a different config file for a different number of classes in order to train.
Thank you.

I have used my own weights and configuration file. But while I am testing with new images the darknet can detect 4 objects where keras can detect only 2.
Do you know the reason behind this ?

I have the following problem when running "demo.py", could anyone tell me how to solve it?
`
WARNING:tensorflow:From C:\Users\10137\Anaconda3\lib\site-packages\tensorflow\python\framework\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Colocations handled automatically by placer.
C:\Users\10137\Anaconda3\lib\site-packages\keras\engine\saving.py:292: UserWarning: No training configuration found in save file: the model was not compiled. Compile it manually.
warnings.warn('No training configuration found in save file: '
dog.jpg
Traceback (most recent call last):

File "", line 1, in
runfile('C:/Users/10137/Desktop/learn/python/yolov3_keras/demo.py', wdir='C:/Users/10137/Desktop/learn/python/yolov3_keras')

File "C:\Users\10137\Anaconda3\lib\site-packages\spyder_kernels\customize\spydercustomize.py", line 786, in runfile
execfile(filename, namespace)

File "C:\Users\10137\Anaconda3\lib\site-packages\spyder_kernels\customize\spydercustomize.py", line 110, in execfile
exec(compile(f.read(), filename, 'exec'), namespace)

File "C:/Users/10137/Desktop/learn/python/yolov3_keras/demo.py", line 152, in
image = detect_image(image, yolo, all_classes)

File "C:/Users/10137/Desktop/learn/python/yolov3_keras/demo.py", line 90, in detect_image
boxes, classes, scores = yolo.predict(pimage, image.shape)

File "C:\Users\10137\Desktop\learn\python\yolov3_keras\model\yolo_model.py", line 212, in predict
outs = self._yolo.predict(image)

File "C:\Users\10137\Anaconda3\lib\site-packages\keras\engine\training.py", line 1169, in predict
steps=steps)

File "C:\Users\10137\Anaconda3\lib\site-packages\keras\engine\training_arrays.py", line 294, in predict_loop
batch_outs = f(ins_batch)

File "C:\Users\10137\Anaconda3\lib\site-packages\keras\backend\tensorflow_backend.py", line 2715, in call
return self._call(inputs)

File "C:\Users\10137\Anaconda3\lib\site-packages\keras\backend\tensorflow_backend.py", line 2675, in _call
fetched = self._callable_fn(*array_vals)

File "C:\Users\10137\Anaconda3\lib\site-packages\tensorflow\python\client\session.py", line 1439, in call
run_metadata_ptr)

File "C:\Users\10137\Anaconda3\lib\site-packages\tensorflow\python\framework\errors_impl.py", line 528, in exit
c_api.TF_GetCode(self.status.status))

AbortedError: Operation received an exception:Status: 3, message: could not create a dilated convolution forward descriptor, in file tensorflow/core/kernels/mkl_conv_ops.cc:1111
[[{{node conv2d_2/convolution}}]]
`

Is anyone met this error ?