fb1b5e09b2
faster and more informative training plots ( #1114 )
...
* faster and more informative training plots
* Update utils.py
Looks good. Needs pep8 linting, I'll do that in PyCharm later once PR is in.
* Update test.py
* Update train.py
f for the tb descriptor lets us plot several batches, i.e. to allow us to change L292 to 'if ni < 3' for 3 examples.
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
2020-04-30 13:37:04 -07:00
c6ea2b58ea
auto-accumulate update
2020-04-28 15:06:33 -07:00
992d8af242
faster hsv augmentation ( #1110 )
...
As per https://github.com/ultralytics/yolov3/issues/1096
2020-04-28 12:59:44 -07:00
15f1343dfc
uncached label removal
2020-04-28 11:07:26 -07:00
2518868508
MemoryEfficientMish()
2020-04-27 13:51:21 -07:00
3aa347a321
add HardSwish()
2020-04-27 13:08:24 -07:00
4a4bfb20de
FLOPS verbose=False
2020-04-26 16:31:57 -07:00
a0a3bab9e6
add Mish() support
2020-04-26 16:31:21 -07:00
754a1b5bf8
reduce merge limit to 3000
2020-04-25 21:15:30 -07:00
82a12e2c8e
docker train update
2020-04-22 11:38:48 -07:00
77b3829d56
check_git_status() to train.py
2020-04-22 11:02:09 -07:00
03c6a2d6fa
cleanup
2020-04-21 12:31:37 -07:00
accce6b565
git status check bug fix
2020-04-18 18:06:11 -07:00
693c06b26c
bug fix issues/1067
2020-04-18 12:07:44 -07:00
bf1061c146
cleanup
2020-04-16 16:12:23 -07:00
510eadcfa5
Apex and 'git pull' suggestions
2020-04-15 12:54:56 -07:00
a49ea80218
update initialize_weights()
2020-04-14 15:58:32 -07:00
ac4c90c817
cleanup
2020-04-14 13:08:00 -07:00
763cdd5ae2
detailed image sizes report
2020-04-14 11:51:19 -07:00
029e137bc2
bug fix
2020-04-14 04:34:40 -07:00
1681249588
cleanup
2020-04-14 04:15:53 -07:00
198a5a591d
code cleanup
2020-04-14 04:15:05 -07:00
25725c8569
bug fix
2020-04-14 03:13:30 -07:00
835b0da68a
new modules and init weights
2020-04-14 01:20:57 -07:00
76fb8d48d4
ng dependence removed from build_targets()
2020-04-13 21:25:03 -07:00
b8574add37
new find_modules() fcn
2020-04-13 17:48:30 -07:00
77e6bdd3c1
FLOPs at 480x640, BN init
2020-04-12 18:44:18 -07:00
efc754a794
add generations arg to kmeans()
2020-04-12 12:49:23 -07:00
a34219a54b
pading from (k-1) // 2 to k // 2
2020-04-11 12:18:54 -07:00
b574f765ce
add warning to plot_results()
2020-04-11 11:04:10 -07:00
58edfc4a84
kaiming weight init
2020-04-11 10:45:33 -07:00
2cf23c4aee
add MixConv2d() layer
2020-04-10 18:58:34 -07:00
4bbfda5cde
hist equalization
2020-04-10 17:29:57 -07:00
9bc3a551d9
histogram equalization added to augmentation
2020-04-10 17:24:49 -07:00
398f8eadec
add thr=0.10 to kmean_anchors()
2020-04-10 16:34:32 -07:00
aa8b1098dd
adapt mosaic to img channel count
2020-04-10 16:28:59 -07:00
6736d7d125
swap cv2.INTER_AREA for cv2.INTER_LINEAR
2020-04-10 12:47:07 -07:00
4120ac3aa6
training updates
2020-04-08 21:01:58 -07:00
f54d28ba63
improve assert no labels found
2020-04-08 09:57:59 -07:00
6c5ecaf805
remove label loading during training
2020-04-07 16:57:22 -07:00
b9b14bef59
scale_img() bug fix
2020-04-07 13:35:47 -07:00
067ee264c0
scale_img() bug fix
2020-04-07 12:59:52 -07:00
68f58f4dec
scale_img() update
2020-04-07 12:51:52 -07:00
26fc4fb018
dataloader default color to imagenet mean 114
2020-04-05 18:02:41 -07:00
4da5c6c114
rect padding to 64, mAP increase 42.7 to 42.9
2020-04-05 16:06:27 -07:00
a657345b45
add FeatureConcat() module
2020-04-05 14:47:41 -07:00
a19b1a3b94
line thickness
2020-04-05 11:10:05 -07:00
6203340888
detect.py multi_label default False
2020-04-05 11:05:49 -07:00
00c1fdd805
add MixConv2d() layer
2020-04-03 20:03:44 -07:00
207c6fcff9
merge NMS full matrix
2020-04-02 18:53:40 -07:00
Josh Veitch-Michaelis
Glenn Jocher
Josh Veitch-Michaelis