-
Notifications
You must be signed in to change notification settings - Fork 152
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Yolov6 low fps on .trt model with fp16 #22
Comments
please select a small size model, current the model is so bigger. |
I trained my model on yolov6s.pt Cuda version 11.4 and cudnn 8.2.4 |
can you show your code? so I can see more. |
which python script do I have to share? |
trt.py |
#trt.py import sys class Predictor(BaseEngine): if name == 'main':
|
the output of |
#trt utils import tensorrt as trt class BaseEngine(object):
def nms(boxes, scores, nms_thr):
def multiclass_nms(boxes, scores, nms_thr, score_thr): def preproc(image, input_size, mean, std, swap=(2, 0, 1)):
_COLORS = np.array( def vis(img, boxes, scores, cls_ids, conf=0.5, class_names=None):
|
def detect_video(self, video_path): ======> from this function i am getting fps 23-27 pred.video() gives overall frames of a video more than 500 |
|
you platform is 1080Ti? |
I am using RTX 3070 8gb |
i think you can set the waitkey more small. you can try set |
Sure, I am eagerly waiting for the updates on the camera test model. |
try this solution |
Thanks, for the solution. FPS increased to 90-98 fps Cuda utilisation is still 354 MB Is there any solution to utilizing full gpu/cuda. |
So, I can increase my fps more. |
opencv frame updates are slower than inference time, in fact your model may run faster. |
Kindly provide me some links or sources to implement multiple batches for video inference. |
Using export.py file. I am able to convert my model from .onnx to .trt
While using the trt.py file.
Cuda utilization is only 354 MB and I m getting fps 23-27 only.
can help me to resolve this issue to increase my fps.
The text was updated successfully, but these errors were encountered: