Commit Graph

187 Commits

Author SHA1 Message Date
Joseph Redmon
a79299d317 NIPS demo 2015-06-10 08:49:02 -07:00
Joseph Redmon
cbc9984a17 NIPS 2015-06-10 00:11:41 -07:00
Joseph Redmon
7fe80a2bb5 sync issues 2015-06-09 13:56:33 -07:00
Joseph Redmon
e63b3a6f91 syncing messed something up 2015-06-09 11:17:46 -07:00
Joseph Redmon
28d5a4a913 more detection stuff 2015-05-31 13:49:50 -07:00
Joseph Redmon
c521f87c9e ? 2015-05-25 11:53:10 -07:00
Joseph Redmon
ff7e03325a detection exp 2015-05-20 10:06:42 -07:00
Joseph Redmon
46e1b263e1 testing other losses 2015-05-15 10:25:05 -07:00
Joseph Redmon
7399dd1af5 something 2015-05-12 12:03:32 -07:00
Joseph Redmon
c39dfd188a Fixed maxpool after change 2015-05-11 14:36:45 -07:00
Joseph Redmon
96b578db63 whoops! 2015-05-11 13:54:58 -07:00
Joseph Redmon
516f019ba6 route handles input images well....ish 2015-05-11 13:46:49 -07:00
Joseph Redmon
dc0d7bb8a8 forgot route layer 2015-05-08 10:37:39 -07:00
Joseph Redmon
0cbfa46461 stuff 2015-05-08 10:33:47 -07:00
Joseph Redmon
e7688a05a1 no idea 2015-05-06 14:08:16 -07:00
Joseph Redmon
372980d690 i don't know what's going on anymore 2015-05-04 11:29:21 -07:00
Joseph Redmon
989ab8c38a IOU loss function 2015-04-24 10:27:50 -07:00
Joseph Redmon
feabcc31de gonna fuck shit up 2015-04-21 12:58:03 -07:00
Joseph Redmon
451ef0a0a6 It's time, to du-du-du-du-DU-DU-DUEL!!
https://www.youtube.com/watch?v=IVmtUK_1jh4
2015-04-20 08:46:12 -07:00
Joseph Redmon
f199fd3b64 per image randomness in crop layer 2015-04-17 12:32:54 -07:00
Joseph Redmon
47528e37cf crop layer scaling and trans on cpu 2015-04-15 01:04:38 -07:00
Joseph Redmon
d97331b88f level adjustment for images 2015-04-15 00:32:32 -07:00
Joseph Redmon
b0106d7bde probably stuff changed 2015-04-13 14:09:55 -07:00
Joseph Redmon
390a0cf923 not much changed... 2015-04-11 01:24:07 -07:00
Joseph Redmon
5a49c1d962 rotation 2015-04-09 23:00:33 -07:00
Joseph Redmon
4f50e29365 big change to images 2015-04-09 15:18:54 -07:00
Joseph Redmon
1fd10265f8 detection better? 2015-04-07 15:25:30 -07:00
Joseph Redmon
a05c4bd2e9 sqrt 2015-04-01 10:25:50 -07:00
Joseph Redmon
81751b47dd ..... and back to coords 2015-03-30 12:04:03 -07:00
Joseph Redmon
ad59d7e68e sides of box instead of coords 2015-03-30 11:09:01 -07:00
Joseph Redmon
6553b3f0e3 no comment 2015-03-29 19:31:47 -07:00
Joseph Redmon
f98bf6bbdb We do our OWN resizing! 2015-03-28 16:11:37 -07:00
Joseph Redmon
cf0300ea55 dropout probably ok 2015-03-27 17:32:01 -07:00
Joseph Redmon
d7d7da2653 Fixed im2col mistake >< face#palm 2015-03-26 19:13:59 -07:00
Joseph Redmon
e92f7d301c smaller gridsize in bias 2015-03-24 18:27:12 -07:00
Joseph Redmon
56b6561ae4 stuff changed probably 2015-03-24 13:20:56 -07:00
Joseph Redmon
7100de0b59 going to break stuff 2015-03-22 21:28:45 -07:00
Joseph Redmon
664c5dd2f2 Subdivisions for batches 2015-03-22 09:56:40 -07:00
Joseph Redmon
9d418102f4 using caffe's im2col, it's so much better\! 2015-03-21 14:17:39 -07:00
Joseph Redmon
4af116e996 gonna change im2col 2015-03-21 12:25:14 -07:00
Joseph Redmon
dcb000b553 refactoring and added DARK ZONE 2015-03-11 22:20:15 -07:00
Joseph Redmon
f047cfff99 renamed sigmoid to logistic 2015-03-08 11:31:12 -07:00
Joseph Redmon
655f636a42 detection layer fixed 2015-03-08 11:25:28 -07:00
Joseph Redmon
2313a8eb54 Split commands into different files 2015-03-06 10:49:03 -08:00
Joseph Redmon
26cddc6f93 going to split things up 2015-03-06 09:49:05 -08:00
Joseph Redmon
fb9e0fe336 Big changes to detection 2015-03-04 14:56:38 -08:00
Joseph Redmon
5f4a5f59b0 captcha stuff 2015-02-23 18:52:05 -08:00
Joseph Redmon
0f645836f1 Detection is back, baby\! 2015-02-10 19:41:03 -08:00
Joseph Redmon
979d02126b Generalizing conv layer so deconv is easier 2015-02-09 13:27:58 -08:00
Joseph Redmon
7ee45082f1 doing some testing w/o rand 2015-02-09 13:24:32 -08:00
Joseph Redmon
2f62fe33c9 saving weight files as binaries, hell yeah 2015-02-06 18:53:53 -08:00
Joseph Redmon
bfffadc755 Stable place to commit 2015-02-04 12:41:20 -08:00
Joseph Redmon
0f1a31648c idk, probably something changed 2015-01-30 22:05:23 -08:00
Joseph Redmon
c592fc7491 New name: Darknet 2015-01-27 13:52:37 -08:00
Joseph Redmon
153705226d Bias updates bug fix 2015-01-27 13:31:06 -08:00
Joseph Redmon
809f924db2 CUDA so fast 2015-01-22 16:38:24 -08:00
Joseph Redmon
4ac78c8926 I am so done with opencl, switching to cuda 2015-01-20 13:26:46 -08:00
Joseph Redmon
6e1d5b45de fast sort of working 2015-01-19 22:06:18 -08:00
Joseph Redmon
08b757a0bf Stable, needs to be way faster 2015-01-14 12:18:57 -08:00
Joseph Redmon
aa5996d58e Faster 2015-01-12 17:27:08 -08:00
Joseph Redmon
f26da0ad5c Need to fix line reads 2014-12-28 09:42:35 -08:00
Joseph Redmon
4ab366a805 some fixes, some other experiments 2014-12-22 14:35:37 -08:00
Joseph Redmon
f88baf4a3a 99 problems 2014-12-18 15:46:45 -08:00
Joseph Redmon
19d3ae7267 message 2014-12-18 13:21:30 -08:00
Joseph Redmon
47914146d9 working stuffs 2014-12-18 11:28:42 -08:00
Joseph Redmon
ecbeec86bf changed nist testing 2014-12-18 11:27:13 -08:00
Joseph Redmon
9e98f1b68e Not sure 2014-12-18 10:47:33 -08:00
Joseph Redmon
d6cdea49eb Small fixes 2014-12-16 15:41:36 -08:00
Joseph Redmon
884045091b lots of cleaning 2014-12-16 15:34:10 -08:00
Joseph Redmon
d6fbe86e7a updates? 2014-12-16 11:40:05 -08:00
Joseph Redmon
90d354a2a5 fixed dropout >< 2014-12-13 12:01:21 -08:00
Joseph Redmon
79fffcce3c Better imagenet distributed training 2014-12-11 13:15:26 -08:00
Joseph Redmon
aea3bceeb1 timing code 2014-12-08 12:11:04 -08:00
Joseph Redmon
cb1f33c6ae Fixed race condition in server 2014-12-08 11:48:57 -08:00
Joseph Redmon
d81ab1942b mnist now 2014-12-07 23:59:45 -08:00
Joseph Redmon
a99050f0c8 Some fixes to momentum 2014-12-07 20:16:21 -08:00
Joseph Redmon
28e2115272 Distributed training 2014-12-07 00:41:26 -08:00
Joseph Redmon
1edcf73a73 Detection good, split up col images 2014-12-03 23:20:29 -08:00
Joseph Redmon
ff67f03476 Starting on server/client 2014-12-03 08:48:07 -08:00
Joseph Redmon
b77a8f3987 stable 2014-11-28 10:38:26 -08:00
Joseph Redmon
e36182cd8c cleaned up data parsing a lot. probably nothing broken? 2014-11-21 15:35:19 -08:00
Joseph Redmon
7c120aef23 stable, dropout on gpu 2014-11-19 14:03:51 -08:00
Joseph Redmon
d407bffde9 checkpoint 2014-11-18 13:51:04 -08:00
Joseph Redmon
b13ad6d5fd Can validate on imagenet now 2014-11-05 14:49:58 -08:00
Joseph Redmon
2b2441313b col2im maybe a little faster 2014-10-30 11:28:37 -07:00
Joseph Redmon
27d0c922ea im2col actually works now 2014-10-29 23:47:50 -07:00
Joseph Redmon
1c0fd9bb47 im2col slightly faster 2014-10-29 23:26:41 -07:00
Joseph Redmon
af4e4f92dc getting rid of sub_arrays, nvidia driver memory leak 2014-10-27 19:45:06 -07:00
Joseph Redmon
edbccdfcaf Maybe something changed? 2014-10-25 22:04:34 -07:00
Joseph Redmon
5c9a773bb6 Maybe some stuff changed 2014-10-25 17:35:25 -07:00
Joseph Redmon
14303717dc Fast, needs to be faster 2014-10-25 11:57:26 -07:00
Joseph Redmon
158bb1bee9 softmax on gpu 2014-10-21 14:49:18 -07:00
Joseph Redmon
9b3c7136f3 Fixing up maxpool layer 2014-10-16 15:17:23 -07:00
Joseph Redmon
7756cccb79 Refactored connected to use blas 2014-10-13 22:31:48 -07:00
Joseph Redmon
787d534560 Convolutional working on GPU 2014-10-13 00:31:10 -07:00
Joseph Redmon
76ee68f96d Trying some stuff w/ dropout 2014-08-27 19:11:46 -07:00
Joseph Redmon
176d65b765 Nist NIN testing multi-crop 2014-08-11 12:52:07 -07:00
Joseph Redmon
7add111509 maxpool fixed, good on mnist 2014-08-09 08:16:37 -07:00
Joseph Redmon
d9f1b0b16e probably how maxpool layers should be 2014-08-08 12:04:15 -07:00
Joseph Redmon
b32a287e38 Merge branch 'master' of pjreddie.com:jnet
Conflicts:
	src/cnn.c
2014-07-17 10:17:52 -07:00