From 88e3884c2d0fa7bbd258055f562c6d02bbf7f64f Mon Sep 17 00:00:00 2001 From: Jia Guo Date: Thu, 25 Apr 2019 15:17:20 +0800 Subject: [PATCH] Update README.md --- iccv19-challenge/README.md | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/iccv19-challenge/README.md b/iccv19-challenge/README.md index b9c12c0..10adb72 100644 --- a/iccv19-challenge/README.md +++ b/iccv19-challenge/README.md @@ -10,6 +10,8 @@ Firstly please read the workshop [homepage](https://ibug.doc.ic.ac.uk/resources/ Test Submission Server[TODO] +================== + **How To Start:** **Training:** @@ -19,6 +21,8 @@ Test Submission Server[TODO] 3. Refer to the `retina` dataset config section in `sample_config.py` and copy it to your own`config.py`. 4. Start training with `CUDA_VISIBLE_DEVICES='0,1,2,3' python -u train.py --dataset retina --network [your-network] --loss arcface`. It will output the accuracy of lfw, cfp_fp and agedb_30 every 2000 batches by default. +------------------ + **Testing:** 1. testdata-image from baiducloud or dropbox. These face images are all pre-processed and aligned so no need to do further modification. @@ -31,19 +35,27 @@ Test Submission Server[TODO] You can also check the verification performance during training time on LFW,CFP_FP,AgeDB_30 datasets. +------------------ + **Evaluation:** Final ranking is determined by accuracy only, for all valid submissions. For example, score of track-1 will be calculated by ``TAR_glint-light+TAR_iqiyi-light`` while ``TAR_glint-large+TAR_iqiyi-large`` for track-2. +------------------ + **Discussion:** [https://github.com/deepinsight/insightface/issues/632](https://github.com/deepinsight/insightface/issues/632) +------------------ + **Baseline:** 1. Network y2(a deeper mobilefacenet): 933M FLOPs. TAR_image: [TODO], TAR_video: [TODO] 2. Network r100fc(ResNet100FC-IR): 24G FLOPs. TAR_image: [TODO], TAR_video: [TODO] +------------------ + **Candidate solutions:** 1. Use slightly deeper or wider mobile-level networks.