RepMLP: Re-parameterizing Convolutions into Fully-connected Layers for Image Recognition
Released the code of RepMLP together with an example of checking the equivalence between a training-time and an inference-time RepMLP.
Just check it by
python repmlp.py
Will be updated in several days.
Light Block is not fast as the paper says
def test(network, p=True):
x = torch.ones(128, 3, 224, 224).cuda()
model = network.cuda()
if p: print(model)
model.eval()
with torch.no_grad():
# warm iters
for i in range(20):
y = model(x)
# inference test
iters = 50
start = time.time()
for i in range(iters):
y = model(x)
end = time.time()
print((end-start)/iters, 's')
print(y.shape)
if __name__ == "__main__":
torch.backends.cudnn.benchmark=True
test(create_RepMLPRes50_Base_224(deploy=True), False)
test(create_RepMLPRes50_Light_224(deploy=True), False)
test(create_RepMLPRes50_Bottleneck_224(deploy=True), False)
with Titan XP
Base: 17.1 ms
Light Block: 16.9 ms
Bottleneck: 18.6 ms
This design of repmlp-resnet is different from the lastest repmlpnet, and it shows great face recognition accuracy.
why not keep repmlp-resnet in this repo?
关于在单位阵上做卷积，单位阵里有很多0啊，局部信息不会丢失嘛，（还是我理解错了） 比如这段代码里： https://github.com/DingXiaoH/RepMLP/blob/55c76774fb915b8cfde3c029fc68a60dfd5d1515/repmlp.py#L107 假设输入就是(1,1,3,3), groups=1, c_in=c_out=1, 就是简单地在一张(3,3)的图上做一个3x3卷积。 I = torch.eye(9).repeat(1,1).reshape(9,1,3,3) I = tensor([[[[1., 0., 0.], [0., 0., 0.], [0., 0., 0.]]], [[[0., 1., 0.], [0., 0., 0.], [0., 0., 0.]]], [[[0., 0., 1.], [0., 0., 0.], [0., 0., 0.]]], [[[0., 0., 0.], [1., 0., 0.], [0., 0., 0.]]], [[[0., 0., 0.], [0., 1., 0.], [0., 0., 0.]]], [[[0., 0., 0.], [0., 0., 1.], [0., 0., 0.]]], [[[0., 0., 0.], [0., 0., 0.], [1., 0., 0.]]], [[[0., 0., 0.], [0., 0., 0.], [0., 1., 0.]]], [[[0., 0., 0.], [0., 0., 0.], [0., 0., 1.]]]])
在这个上面做卷积，I的形状是(9,1,3,3)，每个(3,3)中只有一个值不为0，卷积后reshape回去，也只有对角元上不为0，这样做(9,9)x(9,1)的矩阵乘的话，相当与给(3,3)里的每一个元素乘了一个单独的值，也不是卷积吧。
Thank you very much for proposing an excellent model and sharing it publicly. Also congratulations on the publication of your results in CVPR. Since I want the RepMLP model should be on one-dimensional data, that is, the input is only [B, C, H]. Would like to ask if it is possible to provide a RepMLP model for such one-dimensional data?
https://github.com/DingXiaoH/RepMLP/blob/main/repmlpnet.py#L49
![How to convert the 1D model of RepMLP [B, C, H]](https://avatars.githubusercontent.com/u/56589047?v=4)
def forward(self, inputs):
x = F.adaptive_avg_pool2d(inputs, output_size=(1, 1))
x = self.fc1(x)
according to the paper, it may should be (h, w)?
A facial recognition device is a device that takes an image or a video of a human face and compares it to another image faces in a database. The structure, shape and proportions of the faces are compared during the face recognition steps.
awesome-deep-text-detection-recognition A curated list of awesome deep learning based papers on text detection and recognition. Text Detection Papers
What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis | paper | training and evaluation data | failure cases and cle
Sign Language Recognition Service This is a Sign Language Recognition service utilizing a deep learning model with Long Short-Term Memory to perform s
Mixed supervision for surface-defect detection: from weakly to fully supervised learning [Computers in Industry 2021] Official PyTorch implementation
Handwritten Line Text Recognition using Deep Learning with Tensorflow Description Use Convolutional Recurrent Neural Network to recognize the Handwrit
OpenCV Image to Dataset Converter Slice a single image of Persian digits into mu
SussyImage Converts an image into funny, smaller amongus characters Demo Mona Lisa | Lona Misa (Made up of AmongUs characters) API I've also added an
ocr-table This project aims to extract tables from scanned image PDFs using Optical Character Recognition. Install Requirements Tesseract OCR sudo apt
Scene Text Localization & Recognition Resources Read this institute-wise: English, 简体中文. Read this year-wise: English, 简体中文. Tags: [STL] (Scene Text L
isearch is an OSINT tool on Instagram. Offers a face recognition reverse image search on Instagram profile feed photos.
Takes a user-inputted query and generates a machine learning image recognition model that determines if an inputted image is or isn't their query
Deskew by Marek Mauder https://galfar.vevb.net/deskew https://github.com/galfar/deskew v1.30 2019-06-07 Overview Deskew is a command line tool for des
IMGUR5K Handwriting Dataset To run the code for downloading the urls and generate corresponding annotations : Usage: python download_imgur5k.py --data
OpenCV-ToothPaint3-Advanced-Digital-Image-Editor This application named ‘Tooth Paint’ version TP_2020.3 (64-bit) or version 3 was developed within a w
Script_Convertir_PDF_IMG_TXT Este script de pyhton convierte un pdf en Imagen luego utilizando tesseract como motor OCR convierte la Imagen a Texto. p
Image Thresholding is used for image segmentation. From a grayscale image, thresholding can be used to create binary images. In thresholding we pick a threshold T.
Table of Contents Overview Requirements Demo Modules Overview This python package contains modules to help with finding and extracting tabular data fr
hocr-tools About About the code Installation System-wide with pip System-wide from source virtualenv Available Programs hocr-check -- check the hOCR f
4 Mar 19, 2022
2.4k Jan 8, 2023
3.2k Jan 4, 2023
1 Jan 8, 2022
169 Dec 30, 2022
224 Jan 7, 2023
14 Dec 29, 2022
14 Aug 18, 2022
209 Dec 6, 2022
901 Dec 11, 2022
20 Oct 25, 2022
1 Jan 13, 2022
127 Dec 3, 2022
213 Dec 26, 2022
1 Nov 5, 2021
1 Jan 27, 2022
3 Feb 15, 2022
311 Dec 24, 2022
285 Dec 8, 2022