Mr.Li
Mr.Li
请问本项目在使用vlc播放时候，会有多长的延迟呢
class DamoYoloONNXDetector: def __init__(self, config_path, onnx_model_path, device='cuda:0', infer_size=(640, 640), conf_threshold=0.5): self.config = parse_config(config_path) self.infer_size = infer_size self.conf_threshold = conf_threshold self.device = device if torch.cuda.is_available() and 'cuda' in device else 'cpu'...
请问有人能帮忙回答一下吗 谢谢
我这边试了好像不行，只能使用7B的backbone
> [@wanglebron](https://github.com/wanglebron) > > Definitely not as is since the 7b/16 variant (backbone used for DETR-like detection in the paper) has a different amount of features than ViT-s or ViT-h...