tag:blogger.com,1999:blog-940850498937689846.post369755538682179423..comments2020-12-25T04:04:07.034-05:00Comments on Ankan's Computer Vision Blog: Face Detection with YOLO (You Only Look Once)Ankan Bansalhttp://www.blogger.com/profile/10002322487327076616noreply@blogger.comBlogger9125tag:blogger.com,1999:blog-940850498937689846.post-34369602150877854032017-12-29T13:44:51.001-05:002017-12-29T13:44:51.001-05:00Hi,
First of all, thanks for sharing your experie...Hi,<br /><br />First of all, thanks for sharing your experience.<br />A few questions: <br /><br />1. How long until convergence in the WIDER FACE dataset (batches, epochs) ?<br />2. Fine-tuning on FDDB is just another training on top of the weight file ?<br />2. How long did it take to fine-tune ?<br /><br />regards<br />Fábio<br /> Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-940850498937689846.post-27618932001076874102017-09-18T21:00:27.411-04:002017-09-18T21:00:27.411-04:00We are interested in your workWe are interested in your workAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-940850498937689846.post-43338596638955952652017-09-18T21:00:11.909-04:002017-09-18T21:00:11.909-04:00Please contact us at tech@spotcrowd.comPlease contact us at tech@spotcrowd.comAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-940850498937689846.post-81517079801795270522017-04-13T21:30:49.463-04:002017-04-13T21:30:49.463-04:00Hi Jumabek. I am sorry for replying this late. I s...Hi Jumabek. I am sorry for replying this late. I somehow missed your comment. If you still need them, the codes are here:<br />https://github.com/ankanbansal/fddb-for-yoloAnkan Bansalhttps://www.blogger.com/profile/10002322487327076616noreply@blogger.comtag:blogger.com,1999:blog-940850498937689846.post-20332958149659879252017-03-14T22:51:21.374-04:002017-03-14T22:51:21.374-04:00Hi Thanks for sharing, I am planning to train face...Hi Thanks for sharing, I am planning to train face detector on V2, for now I will convert the Wider to YOLO format using your code. Could you also share FDDB conversion code. If possible both from ellipse to BBox and from Regular BBox to YOLO or maybe from Ellipse to YOLO BBox depending on your scenario. Thanks a bundleAnonymoushttps://www.blogger.com/profile/09726088806949292741noreply@blogger.comtag:blogger.com,1999:blog-940850498937689846.post-45558184975081761342017-03-03T14:43:14.529-05:002017-03-03T14:43:14.529-05:00Hi Sagar,
I will not be able to put up the update...Hi Sagar,<br /><br />I will not be able to put up the updated weights. However, for training on WIDER, I basically had to generate a label file for each image. Each label file has 1 line for each face in the corresponding image. Each line is of the format:<br />class xc yc w h<br />Where xc and yc are the coordinates of the centre of the box normalised by the image width and height respectively. And w and h are the width and height of the box normalised by the image width and height respectively. <br />The code for this is here:<br />https://goo.gl/poSLl6<br /><br />Then you will need to change the train_images and backup_directory variables in src/yolo.c code and some other minor changes like class names and class labels. <br /><br />You can find my config file here:<br />https://goo.gl/C0p0QA<br />Ankan Bansalhttps://www.blogger.com/profile/10002322487327076616noreply@blogger.comtag:blogger.com,1999:blog-940850498937689846.post-51877071885823165712017-03-03T14:21:35.095-05:002017-03-03T14:21:35.095-05:00I trained V1.I trained V1.Ankan Bansalhttps://www.blogger.com/profile/10002322487327076616noreply@blogger.comtag:blogger.com,1999:blog-940850498937689846.post-91390157088798629182017-02-15T20:59:30.204-05:002017-02-15T20:59:30.204-05:00What changes did you make to enable training for W...What changes did you make to enable training for WIDER? Can you please point to the reference code, and if possible the updated weights? Wish to compare the performance against other networksSagar Bhokrehttps://www.blogger.com/profile/01204972421408195645noreply@blogger.comtag:blogger.com,1999:blog-940850498937689846.post-12925408170527932762017-02-13T05:50:11.639-05:002017-02-13T05:50:11.639-05:00Hi, Did you train yolo-v1 or yolo-v2 for the face ...Hi, Did you train yolo-v1 or yolo-v2 for the face detection?shahttps://www.blogger.com/profile/13189691170431367994noreply@blogger.com