joojs / fairface Goto Github PK

Hello,
I would like to know what is the purpose of the column service_testin the labels csv file.
Does it mean that for this pictures the information have been double checked ?

Thank you

Great work guys! Is there any pretrained model in fairface dataset available? @joojs

Thanks for sharing the dataset.

Looking forward to an open data set，THANK YOU A LOT

Hi,
Tnx for the dataset - is it possible to also obtain identity annotations as they do not seem to be included?
Regards

I get only images, any label infromation such as txts?

Hi,
is there a pre-trained model for this?

I am a newbie in computer vision, and if anybody knows how to make a video-clip ethnicity analysing model, please let me know! Thanks in advance.

Hello,
In the Arxiv article you didn't mention the batch size during the training.

To compare model performance of different datasets, weused an identical model architecture, ResNet-34 [18], to betrained from each dataset. We used ADAM optimization[29] with a learning rate of 0.0001. Given an image, we de-tected faces using the dlib1’s CNN-based face detector [28]and ran the attribute classifier on each face. The experimentwas done in pyTorch.

With what batch size did you get the of the paper ?

Thank you

Is there a download link available?

FYI images redistributed from YFCC100M require attribution of the original author's work (licenses are available in the original YFCC100M dataset file).

The YFCC100M authors also recommend using their MD5 hash as the identifiers instead of arbitrary sequential numbering. This will make the CC license lookup easier too.

if you wish to release data that refers to photos/videos from our dataset, please refer to their hashes instead of their identifiers.

Good job! But it seems that the dataset released has no attribute annotations. Do you have any plan of releasing the label? Thank you. @joojs

https://github.com/joeng03/Ethnicity-Recognition-with-fair-face-dataset

Very good job and i am looking forward to your dataset, Thank you!

Congrats on this paper! Do you have an estimate of when the dataset will be released?

The solution works, but with extremely slow performance on the local computer with GPU:
using CUDA?: True
---0/637---
--- Face detection: 17.417674779891968 seconds ---
--- Face detection: 17.244201183319092 seconds ---
--- Face detection: 17.647963762283325 seconds ---

GPU:
NVIDIA GeForce RTX 3060 Ti
NVIDIA-SMI 470.82.00 Driver Version: 470.82.00 CUDA Version: 11.4

On another GPU's in a Google Colab it's takes 10 times faster

Please advise?

I couldn't access the labels file :
train https://drive.google.com/file/d/1i1L3Yqwaio7YSOCj7ftgk8ZZchPG7dmH/view
val https://drive.google.com/file/d/1wOdja-ezstMEp81tX1a-EYkFebev4h7D/view
Did you take them down ?
Did someone else able to access them ?

Hello,

I cannot achieve same scores with your training scores when I use your dataset and the parameters you mentioned in the paper (Resnet34, Adam optimizer with 0.0001 LR). Can you please give more details about your training procedure?