Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Aug 4, 2020

• Set theme jekyll-theme-leap-day

  v-iashin

  committed
  81df245

Commits on Jul 19, 2020

• more enhancements in README

  v-iashin

  committed
  ae76d0f

Commits on Jul 18, 2020

• more enhancements

  v-iashin

  committed
  eb16bba

• more enhancements in readme

  v-iashin

  committed
  9d117b2

• enhanced README + added conda env

  v-iashin

  committed
  9000f18

• changed PROJECT_PATH

  v-iashin

  committed
  e38af88

• code ref, removed experiments

  v-iashin

  committed
  6c91098

Commits on Aug 18, 2019

• a new string for the case with no objects

  v-iashin

  committed
  f88403e

Commits on Aug 4, 2019

• orientation is fixed on the server side; bbox styles now are more appealing

  v-iashin

  committed
  b89df66

Commits on Jul 31, 2019

• prod is done. TODO: add face model to prod as well

  v-iashin

  committed
  50ea9c5

• refreshed the look of bboxes. TODO: change the prod

  v-iashin

  committed
  8af086e

Commits on Jul 29, 2019

• todo: make bboxes more attractive

  v-iashin

  committed
  e912ffb

Commits on Jul 28, 2019

• first results; added save_model, separated YOLO to another layer; added doc somewhere

  v-iashin

  committed
  69f6200

Commits on Jul 27, 2019

• added metrics to track; run some experiments - found a better lr

  v-iashin

  committed
  0c3c3bb

Commits on Jul 26, 2019

• fixed device bug in predict_and_save; added a MWE for checking the predictions and loading the model; changed the cfg file to fit the face localization problem; train loop now seems to work properl…

  Show description for e1a892d
  v-iashin

  committed
  e1a892d

Commits on Jul 24, 2019

• added: show_examples for ds, training loop?

  v-iashin

  committed
  b629a33

Commits on Jul 23, 2019

• refactoring in the production code (checked). Added doc. todo: training loop

  v-iashin

  committed
  b041e3c

Commits on Jul 22, 2019

• tested loss; solved grad problem?; adapted for resize augmentation. TODO: docs and train loop

  v-iashin

  committed
  5b9a58e

Commits on Jul 21, 2019

• refactor the YOLO layer added loss. TOOD: test grads for t_wh + test losss

  v-iashin

  committed
  5ca952a

Commits on Jul 14, 2019

• added loss but not tested. Todo: loss expects t* instead of predictions which are scaled and added to corner coords; add docs

  v-iashin

  committed
  7479c3a

Commits on Jun 27, 2019

• decided to fix IoU though. todo: finish the doc for make_targets + dataset doc

  v-iashin

  committed
  5147721

Commits on Jun 26, 2019

• The problem was in IoU but I think my implementation is right

  v-iashin

  committed
  fea920e

Commits on Jun 25, 2019

• added confidences on the output images that a user sees

  v-iashin

  committed
  249a5c5

Commits on Jun 24, 2019

• inspect differences

  v-iashin

  committed
  6511d14

Commits on Jun 23, 2019

• make_targets is done. todo: test snippet and finish doc

  v-iashin

  committed
  2163524

Commits on Jun 21, 2019

• vectorized IoU is now can be calculated provided with w and h only

  v-iashin

  committed
  6b9622d

Commits on Jun 20, 2019

• Intermediary commit

  v-iashin

  committed
  e8ecaad

Commits on Jun 10, 2019

• wrote <-> for permutations. todo: make_targets and loss

  v-iashin

  committed
  660be15

Commits on Jun 2, 2019

• added dataset for all phases (check). created snippets notebook

  v-iashin

  committed
  6a27183

Commits on May 26, 2019

• meta parsing. preparing things for loss. todo: loss

  v-iashin

  committed
  9f70bcf

Commits on May 12, 2019

• added a readme for frontal page

  v-iashin

  committed
  6f58299

Commits on Apr 28, 2019

• added data. todo: implement loss

  v-iashin

  committed
  efce32b

Commits on Apr 23, 2019

• added README.md

  v-iashin

  committed
  3b5a896

Commits on Apr 22, 2019

• gitignore on dataset

  v-iashin

  committed
  fcf54fa

Commits on Apr 20, 2019

• added docs to main_flask.py

  v-iashin

  committed
  88cd966

Pagination

PreviousNext