我测试得时候图片推理是正常的，但是视频推理就错了，我用你的模型又是正常的，但是视频推理本质就是图片推理吧，所以图片推理正常视频为什么会错，但能想到得问题就是模型问题了，模型最大的问题可能就是没改先验框，所以先验框是必须要修改的？每个模型都需要与自己对应得先验框？还请博主赐教

你好，请问这个包在哪里安装呢

想问一下为什么不是from rknnlite.api import RKNNLite or from rknn.api import RKNN像这样的?还是这意味它包含它们?
还有rknpu和rknnlite之间是什么关系?希望得到你的答案xiexie

I RKNN: [18:07:17.903] RKNN Runtime Information: librknnrt version: 1.5.0 (e6fe0c678@2023-05-25T08:09:20) I RKNN: [18:07:17.903] RKNN Driver Information: version: 0.8.8 I RKNN: [18:07:17.904] RKNN Model Information: version: 4, toolkit version: 1.5.0+1fa95b5c(compiler version: 1.5.0 (e6fe0c678@2023-05-25T16:15:03)), target: RKNPU v2, target platform: rk3588, framework name: PyTorch, framework layout: NCHW, model inference type: static_shape W RKNN: [18:07:17.905] Model initializer tensor data is empty, name: 762_Resize_input1 W RKNN: [18:07:17.905] Model initializer tensor data is empty, name: 762_Resize_input2 W RKNN: [18:07:17.905] Model initializer tensor data is empty, name: 863_Resize_input1 W RKNN: [18:07:17.905] Model initializer tensor data is empty, name: 863_Resize_input2 ./yolov5s_RK3588_i8.rknn done I RKNN: [18:07:18.350] RKNN Runtime Information: librknnrt version: 1.5.0 (e6fe0c678@2023-05-25T08:09:20) I RKNN: [18:07:18.350] RKNN Driver Information: version: 0.8.8 I RKNN: [18:07:18.351] RKNN Model Information: version: 4, toolkit version: 1.5.0+1fa95b5c(compiler version: 1.5.0 (e6fe0c678@2023-05-25T16:15:03)), target: RKNPU v2, target platform: rk3588, framework name: PyTorch, framework layout: NCHW, model inference type: static_shape W RKNN: [18:07:18.351] Model initializer tensor data is empty, name: 762_Resize_input1 W RKNN: [18:07:18.351] Model initializer tensor data is empty, name: 762_Resize_input2 W RKNN: [18:07:18.351] Model initializer tensor data is empty, name: 863_Resize_input1 W RKNN: [18:07:18.351] Model initializer tensor data is empty, name: 863_Resize_input2 ./yolov5s_RK3588_i8.rknn done I RKNN: [18:07:18.764] RKNN Runtime Information: librknnrt version: 1.5.0 (e6fe0c678@2023-05-25T08:09:20) I RKNN: [18:07:18.764] RKNN Driver Information: version: 0.8.8 I RKNN: [18:07:18.765] RKNN Model Information: version: 4, toolkit version: 1.5.0+1fa95b5c(compiler version: 1.5.0 (e6fe0c678@2023-05-25T16:15:03)), target: RKNPU v2, target platform: rk3588, framework name: PyTorch, framework layout: NCHW, model inference type: static_shape W RKNN: [18:07:18.765] Model initializer tensor data is empty, name: 762_Resize_input1 W RKNN: [18:07:18.765] Model initializer tensor data is empty, name: 762_Resize_input2 W RKNN: [18:07:18.765] Model initializer tensor data is empty, name: 863_Resize_input1 W RKNN: [18:07:18.765] Model initializer tensor data is empty, name: 863_Resize_input2 ./yolov5s_RK3588_i8.rknn done W RKNN: [18:07:19.259] Output(1304): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride. W RKNN: [18:07:19.282] Output(1304): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride. W RKNN: [18:07:19.319] Output(1304): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride. W RKNN: [18:07:19.365] Output(1304): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride. Traceback (most recent call last): File "main.py", line 35, in <module> frame, flag = pool.get() File "/home/cat/Code/rknn-multi-threaded/rknnpool.py", line 54, in get return fut.result(), True File "/usr/lib/python3.8/concurrent/futures/_base.py", line 444, in result return self.__get_result() File "/usr/lib/python3.8/concurrent/futures/_base.py", line 389, in __get_result raise self._exception File "/usr/lib/python3.8/concurrent/futures/thread.py", line 57, in run result = self.fn(*self.args, **self.kwargs) File "/home/cat/Code/rknn-multi-threaded/func.py", line 208, in myFunc input1_data = outputs[1].reshape([3, -1]+list(outputs[1].shape[-2:])) IndexError: list index out of range W RKNN: [18:07:19.433] Output(1304): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride.

模型训练用的是从yolov5官方克隆下来的v7.0版本，使用默认export.py 导出torch模型后使用rknn_model_zoo中的model/CV/objection/yolo/RKNN_model_convert文件下的convert.sh文件进行转换

报错原因似乎是size不对？
大佬有空能不能指点一下

在firefly的RK3588开发板上，运行python的demo，帧率只有4，5帧，读取一帧视频流就快200ms了，博主知道咋结局这个问题嘛？

错误信息如下，求大佬指点，用的是项目自带的模型，板子是orange pi 5 plus
E Catch exception when setting inputs.
E Catch exception when setting inputs.
E Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/rknnlite/api/rknn_lite.py", line 200, in inference
self.rknn_runtime.set_inputs(inputs, data_type, data_format, inputs_pass_through=inputs_pass_through)
File "rknnlite/api/rknn_runtime.py", line 1067, in rknnlite.api.rknn_runtime.RKNNRuntime.set_inputs
Exception: The input[0] need 4dims input, but 3dims input buffer feed.

E Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/rknnlite/api/rknn_lite.py", line 200, in inference
self.rknn_runtime.set_inputs(inputs, data_type, data_format, inputs_pass_through=inputs_pass_through)
File "rknnlite/api/rknn_runtime.py", line 1067, in rknnlite.api.rknn_runtime.RKNNRuntime.set_inputs
Exception: The input[0] need 4dims input, but 3dims input buffer feed.

E Catch exception when setting inputs.
E Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/rknnlite/api/rknn_lite.py", line 200, in inference
self.rknn_runtime.set_inputs(inputs, data_type, data_format, inputs_pass_through=inputs_pass_through)
File "rknnlite/api/rknn_runtime.py", line 1067, in rknnlite.api.rknn_runtime.RKNNRuntime.set_inputs
Exception: The input[0] need 4dims input, but 3dims input buffer feed.

E Catch exception when setting inputs.
E Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/rknnlite/api/rknn_lite.py", line 200, in inference
self.rknn_runtime.set_inputs(inputs, data_type, data_format, inputs_pass_through=inputs_pass_through)
File "rknnlite/api/rknn_runtime.py", line 1067, in rknnlite.api.rknn_runtime.RKNNRuntime.set_inputs
Exception: The input[0] need 4dims input, but 3dims input buffer feed.

E Catch exception when setting inputs.
E Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/rknnlite/api/rknn_lite.py", line 200, in inference
self.rknn_runtime.set_inputs(inputs, data_type, data_format, inputs_pass_through=inputs_pass_through)
File "rknnlite/api/rknn_runtime.py", line 1067, in rknnlite.api.rknn_runtime.RKNNRuntime.set_inputs
Exception: The input[0] need 4dims input, but 3dims input buffer feed.

Traceback (most recent call last):
File "/rknn-multi-threaded/main.py", line 35, in
frame, flag = pool.get()
File "/rknn-multi-threaded/rknnpool.py", line 54, in get
return fut.result(), True
File "/usr/lib/python3.10/concurrent/futures/_base.py", line 451, in result
return self.__get_result()
File "/usr/lib/python3.10/concurrent/futures/_base.py", line 403, in __get_result
raise self._exception
File "/usr/lib/python3.10/concurrent/futures/thread.py", line 58, in run
result = self.fn(*self.args, **self.kwargs)
File "/rknn-multi-threaded/func.py", line 207, in myFunc
input0_data = outputs[0].reshape([3, -1]+list(outputs[0].shape[-2:]))
TypeError: 'NoneType' object is not subscriptable
root@orangepi5plus:/rknn-multi-threaded#

在运行build-linux_RK3588.sh时报错
/usr/local/lib/libopencv_stitching.so.4.9.0: error adding symbols: 文件格式错误
修改CMkaelist.txt中OpenCV的路径之后报同样错，OpenCV版本为自己手动安装为4.9.0，请问应该如何修改呢

hi @leafqycc , do you try to add trackers like DeepSort? when I infer my model yolov6 + DeepSort, the final result will lag " TPEs + 1" frames. I print the detetction result and tracking results, then I find the detection result of "TPEs + 2" frame is same as the first frame. This leads to the final error. But I don' t know how to solve it.

您好，我设置并发数量>1 的时候就会出现报错

E RKNN: [06:51:01.403] failed to submit!, op id: 1, op name: Conv:Conv_0, flags: 0x5, task start: 0, task number: 170, run task counter: 0, int status: 0, please try updating to the latest version of the toolkit2 and runtime from: https://eyun.baidu.com/s/3eTDMk6Y (PWD: rknn)

同时查看NPU 使用情况会发现 他并不是使用多个 NPU, 而是单个NPU 100%

请问应该如何解决呢？

Hi, Very nice work and I am impressed by the result on this small RK3588 Sbc. What would be the simplest way to add a png mask.png file to perform the detecting only in the unmaked area and display the final result on the full video?

python推理yolov8s 发现不管是9线程还是6线程 npu占用率始终在80以下，改动了Func后处理部分其他代码均未改动，大佬有大概思路知道问题所在吗

W RKNN: [03:07:58.718] Output(output0): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride.
W RKNN: [03:07:58.718] Output(283): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride.
W RKNN: [03:07:58.718] Output(285): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride.
W RKNN: [03:07:58.751] Output(output0): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride.
W RKNN: [03:07:58.752] Output(283): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride.
W RKNN: [03:07:58.752] Output(285): size_with_stride larger than model origin size, if need run OutputOperator in NPU, please call rknn_create_memory using size_with_stride.

请切换到root用户

CPU定频

echo "CPU0-3 可用频率:"
sudo cat /sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies
sudo echo userspace > /sys/devices/system/cpu/cpufreq/policy0/scaling_governor
sudo echo 1992000 > /sys/devices/system/cpu/cpufreq/policy0/scaling_setspeed
echo "CPU0-3 当前频率:"
sudo cat /sys/devices/system/cpu/cpufreq/policy0/cpuinfo_cur_freq

NPU定频

echo "NPU 可用频率:"
sudo cat /sys/class/devfreq/fde40000.npu/available_frequencies
sudo echo userspace > /sys/class/devfreq/fde40000.npu/governor
sudo echo 900000000 > /sys/class/devfreq/fde40000.npu/userspace/set_freq
echo "NPU 当前频率:"
sudo cat /sys/class/devfreq/fde40000.npu/cur_freq
参照3588的定频脚本，对3568做了以上修改，帧率无变化。
貌似3568不支持多核，在rknnpool.py文件中都修改成了RKNNLite.NPU_CORE_AUTO
现在RKNNLite 为1.6版本，ubuntu20，