在iOS上 运行报错

[goodcyg]

下载是百度网盘里mnn和opencv，在iOS真机上运行报错如下？难道HyperLPR3不支持iOS？

Error for concat size of op [ 866 ], the 2 input not match output
Compute Shape Error for 866
Reshape error: 480 -> 0
Compute Shape Error for /neck/encoder/svtr_block.0/mixer/Reshape_1_output_0
Invalid Tensor, the session may not be ready
Can't run session because not resized

[qianxiaoer]

哈罗，邮件收到了，我会尽快查看，回复，谢谢！

[goodcyg]

iOS中测试代码如下？
-(void)test {
UIImage *image=[UIImage imageNamed:@"999"];
// 将 UIImage 转换为 raw data
NSData *imageData=UIImagePNGRepresentation(image);
uint8_t *data = (uint8_t *)[imageData bytes];
_imageData=imageData;
int width = image.size.width;
int height = image.size.height;

// 创建 Data Buffer
HLPR_ImageData imgData = {0};
imgData.data = data;
imgData.width = width;
imgData.height = height;
imgData.format = STREAM_BGR;
imgData.rotation = CAMERA_ROTATION_0;

HLPR_ContextConfiguration config = {0};
// 设置模型路径
config.models_path =(char *)[self dir].UTF8String;
//config.models_path = "/path/to/your/models";
// 最大识别数量
config.max_num = 5;
// 使用一个线程
config.threads = 1;
// 不使用半精度推理
config.use_half = false;
// 检测框阈值
config.box_conf_threshold = 0.3f;
// 非极大值抑制阈值
config.nms_threshold = 0.4f;
// 识别置信度阈值
config.rec_confidence_threshold = 0.5f;
// 使用高开销检测模式
config.det_level = DETECT_LEVEL_LOW;

_context= HLPR_CreateContext(&config);

HREESULT ret=HLPR_ContextQueryStatus(_context);
if (ret != 0) {
    NSLog(@"Error creating context.");
    return;
}
P_HLPR_DataBuffer buffer = HLPR_CreateDataBuffer(&imgData);
HLPR_DataBufferSetStreamFormat(buffer, imgData.format);

// 更新 Context 并获取识别结果
HLPR_PlateResultList resultList = {0};
HLPR_ContextUpdateStream(_context, buffer, &resultList);

// 显示结果
if (resultList.plate_size > 0) {
    P_HLPR_PlateResult plate = resultList.plates;
    self.resultLabel.text = [NSString stringWithFormat:@"车牌号:%s", plate->code];
} else {
    self.resultLabel.text = @"未识别到车牌号";
}

// 释放 Data Buffer
HLPR_ReleaseDataBuffer(buffer);

}
-(NSString *)dir {
NSString *str2=[[[NSBundle mainBundle] resourcePath] stringByAppendingPathComponent:@"models/r2_mobile"];
NSLog(@"models:%@",str2);
return str2;
}

[goodcyg]

麻烦看一下，我一直卡在这里?多谢

[goodcyg]

我重新整理一下代码，mnn和opencv2都用百度网盘里的，运行demo，测试代码如上test，不报错了，但是返回resultList.plate_size=0? 也就是没有识别出来车牌？