wojciechmo / yolo2 Goto Github PK

I can't find the lrelu function, where is the lrelu function?

Hello, I want to know if yolov2 using this tensorflow model can be used on intel's moveidius?

Hi, you've done some great work here. Just had a question about test.py, you set anchor_h and anchor_w to 960 and 540 respectively. Looking around though, I don't see how you got those numbers. Could you give me some insight?

Thanks

How do you set Negative Exampels (Background Images)?

Hey,
where did you use your lrelu function? the default activation function in tf.layers.conv2d is None.

BR.
Alireza

Hi,

I wish to convert ckpt to pb but need to know the output node name of the file.
And what is the name of input node?

Thanks,

• Does this detection model train totally from scratch with COCO-Dataset or VOC?
  Should we do pre-train with ImageNet?

• GRID_W = 18 & GRID_H = 10
  How to set these two hyper-parameters ?
  (e.g. input image size: 32 * (13, 13) = (416, 416) as described in paper ? ?)

• The anchor file missed

Thanks!

It's really cool !!
Actually, do you have the after-training checkpoint file based on your model??
It really help us to train our own model as the pre-training

I think it's a little strange about the way of roi_w_scale settle (same with roi_h_scale): roi[1] / anchor[1].
Is the roi[1] the ratio of bbox width with original image width or just the raw pixel width?
The grid_w or grid_h is not considered here ?
P.S. Could you share more details about tranning data or skills, I take a try on VOC，but not work very well. Hope more info, Thanks.

Was 'train.py' coding with python2 ?
if end < 0:
you passed None to end, arise TypeError: '<' not supported between instances of 'int' and 'NoneType'

Is there any reason for doing this? I can't think of any.

From the paper and also other keras implementation of YOLO2, I saw normalization of each layer output. Have you implement this layer or just I missed it somewhere, could you point it out. Thank you a lot

Hi thanks for sharing your amazing job!
Just one question:

I am refering to loss method in train.py
is not clear to me why you vectorized the masked_label_c but not masked_pred_c. Should you vectorize the masked_pred_c too?

Thanks

Hello,
I would like to test your source, can you upload the data.csv sample?
Thank you in advance.