-
# Spatial Pyramid Pooling - 空间金字塔池化
针对卷积神经网络重复运算问题,2015年微软研究院的何恺明等提出一种SPP-Net算法,通过在卷积层和全连接层之间加入空间金字塔池化结构(Spatial Pyramid Pooling)代替R-CNN算法在输入卷积神经网络前对各个候选区域进行剪裁、缩放操作使其图像子块尺寸一致的做法。
- 解决了卷积神经网络对图像重…
-
I'm currently using a silabs BLE device emulating the SPP mode. SPP data is transmitted as notified packet with UUID FEC26EC46D7144429F8155BC21D658D6. I can see the data on Android BLE terminal app,…
-
hi i am having a problem with uploading it to my board i am just trying to run the
`
#include
void setup()
{
ps5.begin("1a:2b:3c:01:01:01");
Serial.println("Ready.");
}
void lo…
-
SPP could be improved.
-
博主能把SPP模块也给实现了吗?SPP模块能给模型带来性能提升
-
使用tool下的[generate_tag](https://github.com/chichengcn/gici-open/blob/master/tools/edit_binary/generate_replay_tag/src/generate_tag.cpp)对示例数据重新生成对应的tag(SPP),会出现卡死的情况?
![image](https://github.com/chiche…
-
@AlexeyAB Thank you for your work, I encountered a problem
When I converted to the tensorrt model, I found that the darknet maxpool layer implementation is l-> out_w = (w + l-> pad-l-> size) / l-> st…
ou525 updated
4 years ago
-
Hi @AlexeyAB,
Yolov3-spp performs better on the Coco dataset than Yolov3.
When training with my on images and classes, do you recommend using the Yolov3-spp architecture as well? If so, can I us…
-
I downloaded the Blue Serial app and it works perfectly and displays messages sent from the SPP module correctly. I downloaded code from here and when i try to run it, it wont display the results prop…
-
Hi,
I would like to train yolo3-spp with other resolutions, like 416 and 832 (Default = 608).
In this case, would it be necessary to change the size of the maxpool filters (Default = 5, 9 and 13)?…