Inference on Jetson

[Seeeeeyo]

Hi,

Has anyone run this on a NVIDIA Jetson device (Nano/Xavier/Orin) ?
If so, could you share the inference times you’re seeing, along with which Jetson model, anf which models sizes you used?

Thanks!
Selim

[yichuan1998]

Hi,

Has anyone run this on a NVIDIA Jetson device (Nano/Xavier/Orin) ? If so, could you share the inference times you’re seeing, along with which Jetson model, anf which models sizes you used?

Thanks! Selim

I made it success to run vitpose-s-coco(fp16) on jetson Orin, but it's very slow

[Seeeeeyo]

Hi,
Has anyone run this on a NVIDIA Jetson device (Nano/Xavier/Orin) ? If so, could you share the inference times you’re seeing, along with which Jetson model, anf which models sizes you used?
Thanks! Selim

I made it success to run vitpose-s-coco(fp16) on jetson Orin, but it's very slow

Thanks for your reply! What was the inference time/fps?

[yichuan1998]

Hi,
Has anyone run this on a NVIDIA Jetson device (Nano/Xavier/Orin) ? If so, could you share the inference times you’re seeing, along with which Jetson model, anf which models sizes you used?
Thanks! Selim

I made it success to run vitpose-s-coco(fp16) on jetson Orin, but it's very slow

Thanks for your reply! What was the inference time/fps?

only 3 FPS with code below

import cv2
from easy_ViTPose import VitInference

model = VitInference(
    model="vitpose-s-coco.engine",     
    yolo="yolo11n.engine",              
    model_name="s",
    dataset="coco",
    yolo_size=640,
    is_video=True,      
    single_pose=False,
    device="cuda",
    yolo_step=1
)

cap = cv2.VideoCapture(video_path)   
while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break
    
    rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    kpts_dict = model.inference(rgb)
    vis = model.draw(show_yolo=True)
    
    cv2.imshow("ViTPose", vis)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

model.reset()
cap.release()
cv2.destroyAllWindows()