How to train SqueezeDet by own dataset with one class?

[jwwangchn]

I want to train SqueezeDet use our dataset, this dataset only has one class, but when I use this dataset to train SqueezeDet, it met some error, so I want to know how to train SqueezeDet by own dataset with one class?

[Shashikant255]

I am also facing same issue. Can anyone please let me know how to train SqueezeDet for own dataset with one class?

[jwwangchn]

@Shashikant255 I use two same classes to train SqueezeDet, it works.

[Shashikant255]

@jwwangchn Thanks for your help. I have one more doubt regarding label (bounding box information about object i.e x, y, w, h) file. If any image doesn't have any class or object then its corresponding label file should be empty?

[jwwangchn]

@Shashikant255 As far as I know, you should delete these label files. File reader does not care about images without labels files.

[acmaheri]

You need to change the configuration file, period. Adapt the configuration files such that only one class is specified and make sure no label file is empty and depends on your size of images image resolution (HXW) you have will have to change parameters of anchor boxes.

[Caliber2]

"CLASSES": 1, "CLASS_NAMES": ["my_object"] is this correct way top modify squeeze.config for single object detection? or background needs to be added in CLASS_NAMES and CLASSES = 2?

I am getting no detections of actual object. All detections are in background. Does anybody have a good explanation for this wrong behaviour?