Method evaluate_sort

yolox/evaluators/mot_evaluator.py:223–343

COCO average precision (AP) Evaluation. Iterate inference on the test dataset and the results are evaluated by COCO API. NOTE: This function will change training mode to False, please save states if needed. Args: model : model to evaluate. Retu

(
        self,
        model,
        distributed=False,
        half=False,
        trt_file=None,
        decoder=None,
        test_size=None,
        result_folder=None
    )

        return eval_results

    def evaluate_sort(
        self,
        model,
        distributed=False,
        half=False,
        trt_file=None,
        decoder=None,
        test_size=None,
        result_folder=None
    ):
        """
        COCO average precision (AP) Evaluation. Iterate inference on the test dataset
        and the results are evaluated by COCO API.

        NOTE: This function will change training mode to False, please save states if needed.

        Args:
            model : model to evaluate.

        Returns:
            ap50_95 (float) : COCO AP of IoU=50:95
            ap50 (float) : COCO AP of IoU=50
            summary (sr): summary info of evaluation.
        """
        # TODO half to amp_test
        tensor_type = torch.cuda.HalfTensor if half else torch.cuda.FloatTensor
        model = model.eval()
        if half:
            model = model.half()
        ids = []
        data_list = []
        results = []
        video_names = defaultdict()
        progress_bar = tqdm if is_main_process() else iter

        inference_time = 0
        track_time = 0
        n_samples = len(self.dataloader) - 1

        if trt_file is not None:
            from torch2trt import TRTModule

            model_trt = TRTModule()
            model_trt.load_state_dict(torch.load(trt_file))

            x = torch.ones(1, 3, test_size[0], test_size[1]).cuda()
            model(x)
            model = model_trt
            
        tracker = Sort(self.args.track_thresh)
        
        for cur_iter, (imgs, _, info_imgs, ids) in enumerate(
            progress_bar(self.dataloader)
        ):
            with torch.no_grad():
                # init tracker
                frame_id = info_imgs[2].item()
                video_id = info_imgs[3].item()