FPS of the model

[abhigoku10]

@ArtyZe thanks for the source code and sharing the model . i have few queries

1. what is the difference btw your instance segmeter.c and segmenter.c code , since the mode u share gives output for both the function
2. the time take to predict the on an image is >27 secs which very less , assumption is since its built on c/c++ the fps should be good any ideas on this
3. Any procedure for training on instance segmentation for custom data , please share

Thanks in advance

[ArtyZe]

@abhigoku10 Hello，

1. In this project I realized the semantic segmentation of detect class(background and target), but instance segmentation is as far not works because I think the code of prjeddie is wrong, I have to fix it. Now I have hard work of face detection, so maybe later I can find a intern to finish it;
2. Are you testing it in CPU only? If ture, the time is match what we are expecting for. First, I use the image size 1024×512，But actually you can resize and retrain the net with 214*214, but you have to resize the dataset of train; Second, I use the Network of Unet, but it is not necessary, you can change it reference yolo-tiny; Last, in my real project, I used many ways to reduce the computation, like prune, quantization, parallel compute, I have upload part of demo project to my repo: https://github.com/ArtyZe/yolo_embedded_acceleration After I finish the face detect project, I can update the accelerated segmentation code

[abhigoku10]

@ArtyZe thanks for the response .

1.sure thanks for this update , i shall look into the instance segmentation . 2.yes i am running it on CPU, as suggest i shall re-train it with different resolution . Can i change the network from u-net, yolo-tiny to other network like yolact for which we have to change the cfg and other c files how feasible would it be .. Thanks for sharing the computation time features i shall look into it

[ArtyZe]

@abhigoku10 If you want to change the resolution, what you need to do is only resize train and label image, then change the cfg file.

[abhigoku10]

@ArtyZe sure i shall do it , but any pointers/ references in changing the architecture that would be helpful

[ArtyZe]

@abhigoku10 Are you Chinese? I have written some docs in Chinese. If not, I will update it to readme file of this project this weekend :)

[abhigoku10]

@ArtyZe i am not chinese ,it would be helpful if you could share the readme based on ur time thanku

[abhigoku10]

@ArtyZe i want o use your https://github.com/ArtyZe/yolo_embedded_acceleration code , can you share the weights and cfg files to first test the working of the code