Method predict_image_slice

deploy/python/infer.py:244–369

(self,
                            img_list,
                            slice_size=[640, 640],
                            overlap_ratio=[0.25, 0.25],
                            combine_method='nms',
                            match_threshold=0.6,
                            match_metric='ios',
                            run_benchmark=False,
                            repeats=1,
                            visual=True,
                            save_results=False)

        return self.det_times

    def predict_image_slice(self,
                            img_list,
                            slice_size=[640, 640],
                            overlap_ratio=[0.25, 0.25],
                            combine_method='nms',
                            match_threshold=0.6,
                            match_metric='ios',
                            run_benchmark=False,
                            repeats=1,
                            visual=True,
                            save_results=False):
        # slice infer only support bs=1
        results = []
        try:
            import sahi
            from sahi.slicing import slice_image
        except Exception as e:
            print(
                'sahi not found, plaese install sahi. '
                'for example: `pip install sahi`, see https://github.com/obss/sahi.'
            )
            raise e
        num_classes = len(self.pred_config.labels)
        for i in range(len(img_list)):
            ori_image = img_list[i]
            slice_image_result = sahi.slicing.slice_image(
                image=ori_image,
                slice_height=slice_size[0],
                slice_width=slice_size[1],
                overlap_height_ratio=overlap_ratio[0],
                overlap_width_ratio=overlap_ratio[1])
            sub_img_num = len(slice_image_result)
            merged_bboxs = []
            print('slice to {} sub_samples.', sub_img_num)

            batch_image_list = [
                slice_image_result.images[_ind] for _ind in range(sub_img_num)
            ]
            if run_benchmark:
                # preprocess
                inputs = self.preprocess(batch_image_list)  # warmup
                self.det_times.preprocess_time_s.start()
                inputs = self.preprocess(batch_image_list)
                self.det_times.preprocess_time_s.end()

                # model prediction
                result = self.predict(repeats=50, run_benchmark=True)  # warmup
                self.det_times.inference_time_s.start()
                result = self.predict(repeats=repeats, run_benchmark=True)
                self.det_times.inference_time_s.end(repeats=repeats)

                # postprocess
                result_warmup = self.postprocess(inputs, result)  # warmup
                self.det_times.postprocess_time_s.start()
                result = self.postprocess(inputs, result)
                self.det_times.postprocess_time_s.end()
                self.det_times.img_num += 1