{ "architectures": [ "PPOCRV5MobileDetForObjectDetection" ], "backbone_config": { "block_configs": [ [ [ 3, 16, 32, 1, false ] ], [ [ 3, 32, 32, 2, false ], [ 3, 32, 32, 1, false ] ], [ [ 3, 32, 32, 2, false ], [ 3, 32, 32, 1, false ] ], [ [ 3, 32, 32, 2, false ], [ 5, 32, 32, 1, false ], [ 5, 32, 32, 1, false ], [ 5, 32, 32, 1, false ], [ 5, 32, 32, 1, false ] ], [ [ 5, 32, 32, 2, true ], [ 5, 32, 32, 1, true ], [ 5, 32, 32, 1, false ], [ 5, 32, 32, 1, false ] ] ], "conv_symmetric_num": 4, "depths": [ 1, 2, 2, 5, 4 ], "divisor": 16, "hidden_act": "hardswish", "model_type": "pp_lcnet_v3", "out_features": [ "stage2", "stage3", "stage4", "stage5" ], "out_indices": [ 2, 3, 4, 5 ], "reduction": 4, "scale": 1, "stage_names": [ "stem", "stage1", "stage2", "stage3", "stage4", "stage5" ], "stem_channels": 16, "stem_stride": 2 }, "dtype": "float32", "hidden_act": "hardswish", "id2label": { "0": "text" }, "interpolate_mode": "nearest", "kernel_list": [ 3, 2, 2 ], "label2id": null, "layer_list_out_channels": [ 12, 24, 42, 360 ], "model_type": "pp_ocrv5_mobile_det", "neck_out_channels": 96, "reduction": 4, "transformers_version": "5.16.0.dev0" }