the-house-of-black-and-white / morghulis Goto Github PK

Greetings, first of all i wanna say , that you've done a really nice job, as this framework is incredibly useful. But it seems that there is a problem with converting widerface to darknet format.After executing the export.py file i got only 3501 images and annotations from them, but train dataset of widerface consists of more than 12k images. How can i solve this problem?

https://ufdd.info/

http://www.vision.caltech.edu/html-files/archive.html

AFW dataset is built using Flickr images. It has 205 images with 473 labeled faces. For each face, annotations include a rectangular bounding box, 6 landmarks and the pose angles.

http://www.ics.uci.edu/~xzhu/face/AFW.zip

Hello! Firstly I have to thank you for the amazing work that you have been doing by unifying all major face detection models! While working with morghulis I have found a small bug: In the AFW tensorflow exported the import should be changed "from morghulis import ensure_dir" to from "morghulis.os_utils import ensure_dir". A commit was also submitted for this change.

Hello, i was trying morghulis to convert the widerface dataset to darknet, unfortunately i am not able to get it working, i am calling:

python export.py --dataset widerface --format darknet --data_dir datasets/WIDER --output_dir datasets/WIDER/WIDER_darknet

then this files are created but empty..

do you know what could be going wrong?

greets!

https://www.nist.gov/itl/iad/image-group/ijb-c-dataset-request-form-0

..

http://www.escience.cn/people/geshiming/mafa.html

MALF is the first face detection dataset that supports fine-gained evaluation. MALF consists of 5,250 images and 11,931 faces.

CelebFaces Attributes Dataset (CelebA) is a large-scale face attributes dataset with more than 200K celebrity images, each with 40 attribute annotations.

FDDB dataset contains the annotations for 5,171 faces in a set of 2,845 images.