Commit graph

153 commits

Author SHA1 Message Date
milesial 012fca4715 Removed dense_crf and small fixes
Former-commit-id: de7507ff08510b48e6a0e11da849e0d1c94d3ac8
2019-12-21 22:04:23 +01:00
milesial 5f4ce7dba9 Update mask type for muticlass
Former-commit-id: 4dcb7b8440c5f36ff2565c67f56f8f029b589c80
2019-12-13 17:36:12 +01:00
milesial 4e1f0398a1 Remove assert for multiclass
Former-commit-id: 16809ec7ba14f67bfa5e5a095d3b512ef669a964
2019-12-13 15:58:45 +01:00
milesial 67c4dd34a0 Use classmethod for preprocess
Former-commit-id: ebe3bbb3ca67502db407eaa8273071a871d4b744
2019-12-11 21:57:45 +01:00
milesial 0d1cc25ae2 Convert to tensor in predict.py
Former-commit-id: be1ad32304f2365c0c1db6b0e7fb835acc0fbfed
2019-12-04 12:43:04 +01:00
milesial 5a7e934560 Fix eval.py iterating within batch
Former-commit-id: d4b8040593d4bc82774b1dac647c701b2486c477
2019-12-02 12:06:29 +01:00
milesial 48f45274b7 Update README.md
Former-commit-id: 7cb24a105aa4261eb6daafbb9b0552fd0fdf6fd1
2019-11-23 18:09:00 +01:00
milesial 9d7be6e234 Cleanup + now using tensorboard
Former-commit-id: 79928c84cdf990ef6fe1043a3e4f74b9cc252642
2019-11-23 17:56:14 +01:00
milesial 35f955cbf8 Now using utils.data.Dataset
Former-commit-id: c75d9c075e18add5cd8683faf827937393bf2c94
2019-11-23 14:22:42 +01:00
milesial f5c2771242 Fix predict script (mistake from last commit)
Former-commit-id: 9fd386f5929945f3df1868930d392507fee16a55
2019-11-06 13:23:01 +01:00
milesial 8ed1e09b2a Better support for multiclass
Former-commit-id: 76bebf5f241f579fda7048f5e4a87ee9d49aa423
2019-10-30 19:54:57 +01:00
milesial 6f23624412 Fix typo
Former-commit-id: 72f12645410e87ba18093f055c9d8d9f8045595b
2019-10-30 12:27:03 +01:00
milesial 4c0f0a7a7b Global cleanup, better logging and CLI
Former-commit-id: ff1ac0936c118d129bc8a8014958948d3b3883be
2019-10-26 23:17:48 +02:00
milesial 74f825ce06 Fixed major data loading bug : #46 from pierrezeb/patch-1
Fix split_ids (returned only (id, 0))

The network is now training on both squares and not only the left part of the image

Former-commit-id: e2e46ce509382a45b1db4e1f639aeed568f6cb3e
2019-02-07 13:30:53 +01:00
pierrezeb 543e9e560c fix split_ids (returned only left square!!)
Former-commit-id: a357099e4eace375489f18babb2947d9103733fd
2019-02-07 16:43:47 +09:00
milesial 0f5e7cb0f9 Merge pull request #45 from dmitrysarov/master
evaluation fix : index + 1 in average

Former-commit-id: c8179b82ebe715817890a967b6775ef7875007b3
2019-01-09 14:32:55 +01:00
dmitrysarov 078b03ffa9 index + 1 in average
Former-commit-id: 1dc82b17f075a74f7ef6b20c888835497bcd690b
2019-01-09 15:01:42 +03:00
milesial ef36cad9ba Add note to use python 3 to README
Former-commit-id: fd16a756231b24b5d51810244415a70198492ef9
2018-12-28 17:56:15 +01:00
milesial d3150e8923 Fix upsampling issues (maybe) (#38, #32)
Former-commit-id: 0f45521480d2de8a93407ce43a2a813eefd5d1f6
2018-11-19 11:51:46 +01:00
milesial 5da5e17c08 Merge pull request #37 from rht/master
Move the sigmoid activation to the model itself

Former-commit-id: 7dd7c8b6346033ed1b15a2c1ab54847016a826db
2018-11-11 20:56:15 +01:00
rht 81234be7da Move the sigmoid activation to the model itself
Former-commit-id: e3f8ca7b1ac7c5e9694637a81be260e9b48973b9
2018-11-10 23:11:52 +00:00
milesial 46d1db3115 Another error in diceloss
Former-commit-id: 24a26eaa42b2a8e24335a421047af37cd11adb85
2018-10-28 14:54:14 +01:00
milesial 7e5a8727de Fixed math error in DiceLoss backward pass (#35)
Former-commit-id: 649124b3bd26272bb8442843a546c4e895ec14ff
2018-10-26 18:53:05 +02:00
milesial f2d88f98a7 Switch net to eval mode in predict.py
Former-commit-id: 3af421f3060c30b19ee38bc6e00db2593e53ad62
2018-09-26 08:59:49 +02:00
milesial 341824fb06 Switch net to train mode in train.py
Former-commit-id: ab255c4b3b823dfe8681083aeac0ee78c1a54780
2018-09-26 08:58:49 +02:00
milesial b1ac7151b4 Switch net to eval mode in eval.py
Former-commit-id: 3ab827241455818c7b81172a426d805f9ace0f8f
2018-09-26 08:57:10 +02:00
milesial 3fefc25199 Fixed misuse of epsilon in Dice loss (issue #24)
Former-commit-id: 1e8093de0398364cc65f3c7722b61a0f560df104
2018-08-16 19:50:24 +02:00
milesial f991ffa23e Merge pull request #20 from changjo/patch-1
Fixed argument actions

Former-commit-id: 23e817310461f79888078a2a1ab4b2187a97dc88
2018-07-10 13:17:51 +02:00
Chang Jo Kim 179b08e6a3 Fixed argument actions
Former-commit-id: 7f855d53668a576236b16670fd052d8caf816d80
2018-07-10 11:19:03 +09:00
milesial 14e0ff1c62 Merge pull request #13 from ZiyuanTonyZhang/master
fix scaling issue

Former-commit-id: b282e761f04904697418c9e759d761004a045428
2018-06-23 19:59:38 +02:00
milesial bc465b2d91 Fixed some details on predict + requirements
Former-commit-id: d04263be694bba9dc1cfeb96389a3ad2a4015841
2018-06-23 19:57:53 +02:00
ZiyuanTonyZhang ff85f85dcc fix scaling issue by size change from net, i.e. Issues - error in predict.py #12
Former-commit-id: 4af6e54bd4e1c642282fe512d97aa5a09ce16fca
2018-06-17 02:31:42 -04:00
milesial af08dbb98d Created a requirements.txt
Former-commit-id: ec6b0a4fc245bea8e073fe942cd519e31f1b764f
2018-06-08 19:28:46 +02:00
milesial 02e2314149 Migration to PyTorch 0.4, code cleanup
Former-commit-id: c981801ccc3b74047e94c76e67c4ff1f3097226c
2018-06-08 19:27:32 +02:00
milesial 90e988c10f Fixed bad use of ConvTranspose2D
Former-commit-id: 245b072a82f2e87a9032a1b87136a9930c008afc
2018-06-07 11:20:35 +03:00
milesial 8008b77af6 Removed unused function and general cleanup
Former-commit-id: c34a455f1722e0b899e9e92c7766b83a9a641980
2018-04-09 05:15:24 +02:00
milesial 0da4ad7753 Deleted main.py as it is now outdated
Former-commit-id: f57fa46f2d8cd4b0a13be0ee88708fa4dd4e0a88
2018-04-08 22:36:20 +02:00
milesial 617f334e06 Small changes to README
Former-commit-id: 2c17549a3dc926730e9fcd16ec18610a9e5ec391
2017-11-30 18:50:25 +01:00
milesial 50cbf66b21 Tweaked README
Former-commit-id: f49d66e44456731f0f10cebe8b6d9d2be985fdb2
2017-11-30 08:38:14 +01:00
milesial 6275cb8458 Create LICENSE
Former-commit-id: dbe4c667704454b1e6b816b026d3a3194be120f0
2017-11-30 08:23:15 +01:00
milesial 2d427db832 Cleaned code, added image to README.md
Former-commit-id: 3acf1ff8dadb74e95786fb6ddcf1a90de63f5079
2017-11-30 07:47:43 +01:00
milesial 8b614c3e31 Modified to take any image size (with even width, height > width/2)
Former-commit-id: 2751e6a3df45c1527376a4697d3804d683095d83
2017-11-30 07:19:52 +01:00
milesial 7ea54febec Added CLI for predict, cleaned up code, updated README
Former-commit-id: 77555ccc0925a8fba796ce7e42843d95b6e9dce0
2017-11-30 06:45:19 +01:00
milesial e1bf150da3 Updated README.md
Former-commit-id: 03065569518394e7b686e6995d05599f9822417f
2017-11-30 03:44:29 +01:00
milesial 3c14fc3e63 Final tweaks
Former-commit-id: 547d4580c776afa5782a08fc5288526da82a0972
2017-09-26 21:00:51 +02:00
milesial 932e1c3cf5 Added simple trained model
Former-commit-id: e87526c043c447e5c8d95ee8314fcb7f97c13ce2
2017-08-23 17:38:55 +02:00
milesial 74ee006a7a Added simple predict + submit script 2017-08-21 18:00:07 +02:00
milesial fa40396fff Added simple eval and test CRF 2017-08-19 10:59:51 +02:00
milesial 4063565295 Created a basic train loop + changed a bit loss and utils 2017-08-17 21:16:19 +02:00
milesial 8332f891c3 Changed from deconv to bilinear for upsampling 2017-08-17 15:33:47 +02:00