喜欢用mnn的朋友,这里有一套mnn的适配代码,近期会开源到仓库上,效果如下: 本文版权属于GiantPandaCV,未经允许请勿转载 前言:毕设的一部分,前段时间,在yolov5上进行一系列的消融实验,让他更轻(Flops更小,内存占用更低,参数更少),更快(加入shuffle channel,yolov5 head进行通道裁剪,在320的input_size至少能在树莓派...
|──────ncnn-int8: |Baidu Drive|Google Drive| |──────mnn-fp16: |Baidu Drive|Google Drive| |──────mnn-int4: |Baidu Drive|Google Drive| └──────tengine-fp32: |Baidu Drive|Google Drive| openvino-fp16Baidu DriveGoogle Drive YOLOv5—Liteg.pt: |Baidu Drive|Goo...
mnn for arm-cpu openvino x86-cpu or x86-vpu tensorrt(C++) for arm-gpu or arm-npu or x86-gpu tensorrt(Python) for arm-gpu or arm-npu or x86-gpu Android for arm-cpu Android_demo This is a Redmi phone, the processor is Snapdragon 730G, and yolov5-lite is used for detection. The ...
🛠 A lite C++ toolkit of awesome AI models, support ONNXRuntime, MNN, TNN, NCNN and TensorRT. - add YOLOv5 6.1 ONNXRuntime/MNN C++ · Issue #299 · DefTruth/lite.ai.toolkit
| | |-- mnn | | |-- ncnn | | |-- ort | | |-- tengine | | `-- tensorrt | |-- data | | |-- argoverse_hd.yaml | | |-- coco.yaml | | |-- coco128.yaml | | |-- hyp.finetune.yaml | | |-- hyp.scratch.yaml ...
|──────mnn-int4: |Baidu Drive|Google Drive| └──────tengine-fp32: |Baidu Drive|Google Drive| YOLOv5—Litec.pt:Baidu Drive|Google Drive| └──────openvino-fp16: |Baidu Drive|Google Drive| YOLOv5—Liteg.pt: |Baidu Drive|Google Drive| ...
🛠 A lite C++ toolkit of awesome AI models with ONNXRuntime, NCNN, MNN and TNN. YOLOX, YOLOP, YOLOv6, YOLOR, MODNet, YOLOX, YOLOv7, YOLOv5. MNN, NCNN, TNN, ONNXRuntime, CPU/GPU. - Hanskey802/lite.ai.toolkit
mnnfor arm-cpu openvinox86-cpu or x86-vpu tensorrt(C++)for arm-gpu or arm-npu or x86-gpu tensorrt(Python)for arm-gpu or arm-npu or x86-gpu Androidfor arm-cpu Android_demo This is a Redmi phone, the processor is Snapdragon 730G, and yolov5-lite is used for detection. The performanc...
|──────mnn-int4: |Baidu Drive|Google Drive| |──────onnx-fp32: |Baidu Drive|Google Drive| └──────tengine-fp32: |Baidu Drive|Google Drive| v5lite-c.pt:Baidu Drive|Google Drive| |──────onnx-fp32: |Baidu Drive|Google Drive| ...
v5Lite-e-fp32.mnn3.0mshufflenetv2v5Litee-headmnnArm-cpu v5Lite-e-fp32.tnnmodel v5Lite-e-fp32.tnnproto2.9mshufflenetv2v5Litee-headtnnarm-cpu v5Lite-e-320.onnx3.1mshufflenetv2v5Litee-headonnxruntimex86-cpu @v5lite-s: ModelSizeBackboneHeadFrameworkDesign for ...