search

AlexeyAB / yolo2_light

Light version of convolutional neural network Yolo v3 & v2 for objects detection with a minimum of dependencies (INT8-inference, BIT1-XNOR-inference)
MIT License
301 stars 116 forks source link

Run Yolo2_light for multiple images #16

Open mrkstt opened 5 years ago

mrkstt commented 5 years ago

Hi Alexey,

It's interesting to explore more about this repo. I found it's very efficient. Is it possible to run this version with list of images as an input (as you mentioned in https://github.com/pjreddie/darknet/issues/903). I have image sequences and want to execute it using a single command and output the result as text file.

In addition, would that be possible to show the class name for each bounding box?, e.g., car, bike, etc. (as original Yolo).

You reply is greatly appreciated.

Thanks.

AlexeyAB commented 5 years ago

@mrkstt Hi,

I added some fixes. Update your code from GitHub.

Is it possible to run this version with list of images as an input (as you mentioned in pjreddie/darknet#903).

Yes, now you can use such command with this repository: ./darknet detector test coco.names yolov3.cfg yolov3.weights -thresh 0.24 -dont_show < list.txt > result.txt

In addition, would that be possible to show the class name for each bounding box?, e.g., car, bike, etc. (as original Yolo).

Currently it is implemented for detection only on video ./darknet detector demo. May be I will add it for images.

mrkstt commented 5 years ago

Hi @AlexeyAB ,

Thanks for reply. I listed the images in list.txt as shown below:

input_roi/1.jpg input_roi/2.jpg

But in result.txt only shown this:


Enter Image Path: input_roi/1.jpg: Predicted in 0.054363 seconds.
car: 99%    (left_x:  262   top_y:  534   width:  177   height:  152)
car: 95%    (left_x:  541   top_y:  200   width:   66   height:   65)
motorbike: 85%  (left_x:  566   top_y:  302   width:   38   height:   45)
person: 78% (left_x:  571   top_y:  274   width:   34   height:   45)
motorbike: 39%  (left_x:  572   top_y:  298   width:   30   height:   26)
car: 96%    (left_x:  577   top_y:   97   width:   40   height:   36)
car: 92%    (left_x:  589   top_y:   71   width:   31   height:   26)
car: 92%    (left_x:  621   top_y:   51   width:   33   height:   21)
car: 53%    (left_x:  625   top_y:   27   width:   24   height:   15)
car: 53%    (left_x:  625   top_y:   38   width:   25   height:   14)
car: 38%    (left_x:  639   top_y:   26   width:   22   height:   13)
car: 27%    (left_x:  641   top_y:   16   width:   21   height:   11)
car: 97%    (left_x:  654   top_y:  112   width:   46   height:   48)
car: 75%    (left_x:  675   top_y:   67   width:   35   height:   30)
car: 100%   (left_x:  706   top_y:  226   width:   83   height:  101)
person: 26% (left_x:  738   top_y:   89   width:   12   height:   23)
motorbike: 46%  (left_x:  822   top_y:  156   width:   39   height:   47)
person: 74% (left_x:  840   top_y:  548   width:   46   height:   54)
motorbike: 33%  (left_x:  841   top_y:  590   width:   46   height:   50)
bus: 99%    (left_x:  883   top_y:  165   width:  348   height:  185)
car: 99%    (left_x:  973   top_y:  339   width:  136   height:  125)
car: 78%    (left_x: 1152   top_y:  458   width:   84   height:  171)
car: 51%    (left_x: 1155   top_y:  364   width:   76   height:   95)
Enter Image Path:

It didn't read the next image (which is 2.jpg), only proceed the first image on file.

Am I doing wrong on list.txt?

AlexeyAB commented 5 years ago

I can do it successfully: image