GeneralJing

Results 53 comments of GeneralJing

on iphone7 the inference time is about 90ms, This is not real time, right?

@PierrotLL can you supply some test examples? I met the same question. below is my code: // KalmanFilter f = alloc_filter(2, 1); KalmanFilter f = alloc_filter(7, 4); /* The train...

NVIDIA GeForce RTX 3090 Model: Swin-tiny TRT model path: ./weights/swin_tiny_patch4_window7_224_batch32_fp32.engine [11/14/2023-08:42:58] [TRT] [I] Loaded engine size: 110 MiB [11/14/2023-08:42:58] [TRT] [I] [MemUsageChange] TensorRT-managed allocation in engine deserialization: CPU +0, GPU...

06-22 03:28:14.057 1/1/0 E zlog alita_user_manager.cpp 283: get login state error, username is empty. 06-22 03:28:14.057 1/1/0 V zlog alita_login_manager.cpp 308: GetLoginState:0 06-22 03:28:14.057 1/1/0 V zlog alita_nodejs_bridge_zidl.cpp 458: Write...

嗯,我在仿照mnn的官方demo弄一个,资料比较少。

也想知道,你解惑了吗

一样的问题,训练完检测效果还不错,但是日志loss非常大,iou_loss很大,而且是随着训练的进行loss一直在增加,不知道是什么原因,而且另外一个尺度的各个指标输出都为0

嗯 好的 太感谢了,待会设置下再去试试。你在自己的项目中应用这种蒸馏方式了吗?

大佬,能加个微信吗?后面有啥问题可以向你请教下

+--------+-------+-------+------+ | Scope | mIoU | mAcc | aAcc | +--------+-------+-------+------+ | global | 21.79 | 25.63 | 81.5 | +--------+-------+-------+------+ 这个是我当时训练输出的数据,我看你刚才贴那个后面的两个loss也很大呢loss_cwd这个