yolov8-obb模型转onnx之后再转rknn时量化出错，求大佬解惑

[mozeqiu123]

1，模型用yolov8obb重新训练了自己的模型，只做一个类别的识别。模型不管是bs=1，还是16训练出来的，都会出现这个问题
2，成功转成onnx模型
3，在转rknn时出错如下
4，rknn_model_zoo是git的最新版，我没搞明白为啥会出现这种维度错误，我在ret = rknn.build(do_quantization=do_quant, dataset=DATASET_PATH,rknn_batch_size=1)设置了 bs=1 但是也没用。

`rknn_model_zoo/examples/yolov8_obb/python# python convert.py best.onnx rk3588
I rknn-toolkit2 version: 2.3.0
--> Config model
done
--> Loading model
I Loading : 100%|██████████████████████████████████████████████| 156/156 [00:00<00:00, 75294.76it/s]
done
--> Building model
I OpFusing 0: 100%|██████████████████████████████████████████████| 100/100 [00:00<00:00, 310.62it/s]
I OpFusing 1 : 100%|█████████████████████████████████████████████| 100/100 [00:00<00:00, 186.97it/s]
I OpFusing 0 : 100%|█████████████████████████████████████████████| 100/100 [00:00<00:00, 127.04it/s]
I OpFusing 1 : 100%|█████████████████████████████████████████████| 100/100 [00:00<00:00, 116.26it/s]
I OpFusing 0 : 100%|██████████████████████████████████████████████| 100/100 [00:01<00:00, 97.57it/s]
I OpFusing 1 : 100%|██████████████████████████████████████████████| 100/100 [00:01<00:00, 94.66it/s]
I OpFusing 2 : 100%|██████████████████████████████████████████████| 100/100 [00:06<00:00, 14.70it/s]
I GraphPreparing : 100%|████████████████████████████████████████| 210/210 [00:00<00:00, 9456.87it/s]
I Quantizating 1/4: 0%| | 0/210 [00:00<?, ?it/s]E build: The input('/rknn_model_zoo/datasets/COCO/testimg/001.jpg') shape (1, 640, 640, 3) is wrong, expect 'nhwc' like (16, 640, 640, 3)!
I ===================== WARN(0) =====================
E rknn-toolkit2 version: 2.3.0
I Quantizating 1/4: 0%| | 0/210 [00:00<?, ?it/s]
E build: Traceback (most recent call last):
File "rknn/api/rknn_log.py", line 344, in rknn.api.rknn_log.error_catch_decorator.error_catch_wrapper
File "rknn/api/rknn_base.py", line 1971, in rknn.api.rknn_base.RKNNBase.build
File "rknn/api/rknn_base.py", line 175, in rknn.api.rknn_base.RKNNBase._quantize
File "rknn/api/quantizer.py", line 1417, in rknn.api.quantizer.Quantizer.run
File "rknn/api/quantizer.py", line 899, in rknn.api.quantizer.Quantizer._get_layer_range
File "rknn/api/rknn_utils.py", line 328, in rknn.api.rknn_utils.get_input_img
File "rknn/api/rknn_log.py", line 95, in rknn.api.rknn_log.RKNNLog.e
ValueError: The input('rknn_model_zoo/datasets/COCO/testimg/001.jpg') shape (1, 640, 640, 3) is wrong, expect 'nhwc' like (16, 640, 640, 3)!

I ===================== WARN(0) =====================
E rknn-toolkit2 version: 2.3.0
Traceback (most recent call last):
File "rknn/api/rknn_log.py", line 344, in rknn.api.rknn_log.error_catch_decorator.error_catch_wrapper
File "rknn/api/rknn_base.py", line 1971, in rknn.api.rknn_base.RKNNBase.build
File "rknn/api/rknn_base.py", line 175, in rknn.api.rknn_base.RKNNBase._quantize
File "rknn/api/quantizer.py", line 1417, in rknn.api.quantizer.Quantizer.run
File "rknn/api/quantizer.py", line 899, in rknn.api.quantizer.Quantizer._get_layer_range
File "rknn/api/rknn_utils.py", line 328, in rknn.api.rknn_utils.get_input_img
File "rknn/api/rknn_log.py", line 95, in rknn.api.rknn_log.RKNNLog.e
ValueError: The input('rknn_model_zoo/datasets/COCO/testimg/001.jpg') shape (1, 640, 640, 3) is wrong, expect 'nhwc' like (16, 640, 640, 3)!

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "convert.py", line 60, in
ret = rknn.build(do_quantization=do_quant, dataset=DATASET_PATH)
File "/root/miniforge3/envs/RKNN-Toolkit2/lib/python3.8/site-packages/rknn/api/rknn.py", line 192, in build
return self.rknn_base.build(do_quantization=do_quantization, dataset=dataset, expand_batch_size=rknn_batch_size)
File "rknn/api/rknn_log.py", line 349, in rknn.api.rknn_log.error_catch_decorator.error_catch_wrapper
File "rknn/api/rknn_log.py", line 95, in rknn.api.rknn_log.RKNNLog.e
ValueError: Traceback (most recent call last):
File "rknn/api/rknn_log.py", line 344, in rknn.api.rknn_log.error_catch_decorator.error_catch_wrapper
File "rknn/api/rknn_base.py", line 1971, in rknn.api.rknn_base.RKNNBase.build
File "rknn/api/rknn_base.py", line 175, in rknn.api.rknn_base.RKNNBase._quantize
File "rknn/api/quantizer.py", line 1417, in rknn.api.quantizer.Quantizer.run
File "rknn/api/quantizer.py", line 899, in rknn.api.quantizer.Quantizer._get_layer_range
File "rknn/api/rknn_utils.py", line 328, in rknn.api.rknn_utils.get_input_img
File "rknn/api/rknn_log.py", line 95, in rknn.api.rknn_log.RKNNLog.e
ValueError: The input('rknn_model_zoo/datasets/COCO/testimg/001.jpg') shape (1, 640, 640, 3) is wrong, expect 'nhwc' like (16, 640, 640, 3)!`

[mozeqiu123]

我是按照这里面的方法转的onnx模型，成功转出来了
https://github.com/airockchip/ultralytics_yolov8/blob/main/RKOPT_README.zh-CN.md
输出信息如下，我看到了原始维度就是这个 (16, 3, 640, 640)，是不是要在转换的时候调整输入维度到 (1, 3, 640, 640)再转呀
` python .\exporter.py
Ultralytics 8.3.27 🚀 Python-3.10.13 torch-2.5.1+cpu CPU (13th Gen Intel Core(TM) i9-13900KF)
YOLOv8n-obb summary (fused): 187 layers, 3,080,144 parameters, 0 gradients, 8.3 GFLOPs

PyTorch: starting from 'yolov8n-obb.pt' with input shape (16, 3, 640, 640) BCHW and output shape(s) (16, 20, 8400) (6.3 MB)

ONNX: starting export with onnx 1.14.0 opset 19...
ONNX: slimming with onnxslim 0.1.47...
ONNX: export success ✅ 1.2s, saved as 'yolov8n-obb.onnx' (11.9 MB)

Export complete (2.3s)
Results saved to D:\lilin\ultralytics_yolov8\ultralytics\engine
Predict: yolo predict task=obb model=yolov8n-obb.onnx imgsz=640
Validate: yolo val task=obb model=yolov8n-obb.onnx imgsz=640 data=runs/DOTAv1.0-ms.yaml
Visualize: https://netron.app
`
有没有大佬看看是什么问题呀

[fatcatkk]

我之前跟你遇到了同样的问题，你看看你的default.yaml的# Export settings是不是把format改了，我之前改成了onnx然后就出这个错误了

[mozeqiu123]

我之前跟你遇到了同样的问题，你看看你的default.yaml的# Export settings是不是把format改了，我之前改成了onnx然后就出这个错误了

我是根据这里说的转onnx模型的。只改了自己的模型路径。后面我发现在转换rknn的时候，不进行量化就能转出来，目前正在测试这个转出的rknn模型能否使用。