You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
liuqiliqi 20aa7a9da9 2nd commit 1 year ago
..
atss_0506_lossv2.cpp 2nd commit 1 year ago
atss_0506_lossv2.h 2nd commit 1 year ago
atss_0506_lossv2_config.h 2nd commit 1 year ago
function.h 2nd commit 1 year ago
pool2d.h 2nd commit 1 year ago
sliding_window_unit.h 2nd commit 1 year ago
stream_tools.h 2nd commit 1 year ago

该项目使用了多种优化,在Ultra96v2这样一块极小的开发板上实现了深度可分离卷积,可以对640*360的图像做目标检测,检测种类可达十四种,在图像复杂,干扰性极大的大疆52500检测图像组之中,正确率可达70%以上,同时功耗不超过5W,可以适用于无人机拍摄、高速路边监控、人员设备搜索等等一系列场景之中,同时,由于设备使用资源极少,板子占地面积极小,可以被广泛应用于需要图像识别的各种物联网设备之中。 相比于其他的图像识别模型,我们对整个模型做了从软件到硬件全方位的优化,相同情况下各性能指标均极佳

VHDL Coq Verilog C SystemVerilog other

Contributors (1)