00:21:40 Jerry M: yes 00:21:56 Jerry M: The capture points needed to be adjusted 00:22:08 Jerry M: yes 00:22:17 Jennifer Woodward-Greene: can you scroll up so we can see the operations on each image? 00:25:13 Jerry M: I did use a threshold. There is a bit of a light gradient in this image. 00:25:32 Jerry M: yes 00:54:24 Maximilian Feldman: Is it absolutely critical to have image dimensions of 224 x 224 to use VGG16? What should I do if my input images are 1900 x 1900? 00:57:03 Lucas Heintzman: Does the imagery have to RGB or can CIR be used...so long as the input has 3 bands? 00:59:43 Kerrie Geil: How many of existing CNNs are available for transfer learning and where do we find them? Is there a centralized location where people post/store them? 01:05:20 Andrew.French: keras website says load_img puts into PIL format? 01:05:57 Jerry M: python image library 01:06:28 Jennifer Woodward-Greene: BGR 01:07:52 Lucas Heintzman: BGR is akin to "legacy" Landsat Satellite assignments of: Bands 1 = Blue , Band 2= Green, Band 3 = Red 01:15:15 Maria Laura Cangiano: plate_rack (13.09%) soap_dispenser (12.78%) washbasin (5.91%) 01:15:22 Maria Laura Cangiano: it's a clover 01:15:54 Lucas Heintzman: Hi got a warning that I am trying to DOS attack the system! 01:16:54 Lucas Heintzman: NOpe. 01:16:54 Jennifer Woodward-Greene: From an image of a man holding a goat: 01:16:56 Jennifer Woodward-Greene: oxcart (47.31%) ox (9.61%) yurt (6.77%) 01:17:20 Jerry M: From a watermelon: croquet_ball (16.38%) golf_ball (7.84%) tennis_ball (7.40%) 01:17:31 Jennifer Woodward-Greene: Yum 01:17:40 Elizabeth Chin: from a bowl of oatmeal: mixing_bowl (32.23%) wooden_spoon (17.61%) ice_cream (13.61%) 01:18:02 Andrew.French: from lettuce I got windowscreen timerrwolf and howler monkey 01:18:07 zhanyou xu: great_white_shark (31.49%) albatross (25.21%) tiger_shark (19.21%) for a dolphine 01:18:07 ARS - Kossi Nouwakpo: bike seat --> chain_saw 21% 01:18:26 Gerardo A Armendariz: image of a bear: 01:18:28 Gerardo A Armendariz: brown_bear (99.83%) American_black_bear (0.15%) sloth_bear (0.01%) 01:18:33 Timothy: African_crocodile (43.03%) alligator_lizard (21.00%) American_alligator (13.34%) 01:18:37 Gerardo A Armendariz: yes 01:18:39 Gerardo A Armendariz: it was 01:19:03 Timothy: alligator 01:19:23 Lucas Heintzman: I put a picture of myself in: sarong (7.76%) greenhouse (6.72%) thatch (4.74%) 01:19:57 Kevin Cole: red delicious apple = pomegranate 70 % 01:21:04 ARS - Kossi Nouwakpo: dog -> various breeds of dog 01:22:01 Kerrie Geil: same with the dog 01:22:13 Andrew Russ: Bald Eagle 99.8% Bald Eagle 0.16% Vulture 01:22:33 zhanyou xu: suit (41.35%) Windsor_tie (17.34%) academic_gown (14.34%) for president Obama 01:22:58 zhanyou xu: yes 01:31:10 Jennifer Woodward-Greene: Can I assume this is using multiprocessing? 01:31:51 Jennifer Woodward-Greene: Can you force it to do MP 01:33:46 Jennifer Woodward-Greene: What do you mean it 'found' images... how are they selected? 01:34:14 Jennifer Woodward-Greene: Okay... that's the total! 01:35:02 zhanyou xu: FileNotFoundError: [WinError 3] The system cannot find the path specified: '101_ObjectCategories' 01:35:48 Jennifer Woodward-Greene: WARNING:tensorflow:From :2: Model.fit_generator (from tensorflow.python.keras.engine.training) is deprecated and will be removed in a future version. Instructions for updating: Please use Model.fit, which supports generators. 01:36:59 ARS - Kossi Nouwakpo: It's going to take 3h on my machine. How do you specify gpu processing in keras? 01:37:12 Amy Hudson (she/her): 1hr on HPC! 01:37:28 Jennifer Woodward-Greene: 1 hour my pc! 01:37:49 Kerrie Geil: this morning it took 15 minutes on the brief-low HPC partition 01:39:49 Andrew.French: didn't get far, dimension error: 01:39:52 Andrew.French: ValueError: Error when checking target: expected dense_3 to have shape (101,) but got array with shape (102,) 01:40:30 Kerrie Geil: the Ceres HPC has 1 private GPU that you can access with the gpu-low partition if it's not in use. The new Atlas HPC I believe has multiple community GPUs 01:41:23 Elizabeth Chin: Andrew, did you remove the BACKGROUND_Google/ directory? 01:44:11 Andrew.French: yes found that too and removed, thanks, now only 18 hours till done 01:46:44 Jennifer Woodward-Greene: can you demo the bigger ones if they are too large to fit the day Wednesday? 01:46:58 Jerry M: What types of problems are best suited to transfer learning? 01:50:10 Elizabeth Chin: are models exported through keras exclusive to keras or could they be used through pytorch? 01:55:12 Elizabeth Chin: ok thanks! 01:55:52 Jennifer Woodward-Greene: Excellent, thank you! 01:56:12 Lucas Heintzman: ty 01:56:20 zhanyou xu: thank you all. Bye