Skip to content
This repository has been archived by the owner on Jan 4, 2024. It is now read-only.

ISP errors occurring when set input_size with 416*416 in C++ code #24

Open
STRIVESS opened this issue Dec 25, 2023 · 0 comments
Open

ISP errors occurring when set input_size with 416*416 in C++ code #24

STRIVESS opened this issue Dec 25, 2023 · 0 comments

Comments

@STRIVESS
Copy link

Hello! I follow these instructions in dianjixz/v831_yolo and set both train_size and val_size parameters are 416*416 and to train in my computer, and then I convert my mode into onnx mode and ncnn mode, finally I upload them on the toolbox website to convert into awnn mode.
`

if args.multi_scale:
    print('use the multi-scale trick ...')
    train_size = [640, 640]
    val_size = [416, 416]
else:
    train_size = [416, 416]
    val_size = [416, 416]

`
2023-12-25_15-14

I am sure that my training and modes are correct, and I modify below these codes and replace my trained mode, I try my best to run many time on V831 with C++ code, however there are some ISP errors occurring and the screen is full green(I upload the yolov2_camera_error.log and you can check it), how do I sovle it?
However I follow the same setting in Python code, it can run normally with low frame rate. Does the V831 not support the input_size with 416*416 in C++ code? I am urgent to apply these into my projects recently, thank you for your help!
image
2023-12-25_15-27

These errors below:
image
image

uint32_t res_w = 224, res_h = 224;

char* labels[] = {"aeroplane","bicycle","bird","boat","bottle","bus","car","cat","chair","cow","diningtable","dog","horse","motorbike","person","pottedplant","sheep","sofa","train","tvmonitor"};


float anchors[10] = {0.4165, 0.693 , 0.9765, 1.6065, 1.5855, 3.122 , 2.821 , 1.8515,4.72, 6.26};

char * mud_path = "/root/mud/v831_yolo_voc.mud";

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant