Fine tune retinanet on target animal dataset

[dupeljan]

Thread for responsibility discussion. I think we should launch fine-tuning together, because it would be simpler together. I'll fix the comments of #1 and we can try to launch finetuning

[dupeljan]

@kprokofi Is it complex to add dataset to training pipeline? If it is, I suggest to make a call in pamagiti and to do it together

[dupeljan]

I trained 50 epoch on imagenet subset and got metrics listed below: Pascal VOC AP evaluation 31: toucan: AP 0.9574, precision 0.2416, recall 0.9789 7: water_ouzel: AP 0.9536, precision 0.4218, recall 0.9780 44: redshank: AP 0.9506, precision 0.3932, recall 0.9787 83: three-toed_sloth: AP 0.9174, precision 0.2229, recall 0.9750 11: tailed_frog: AP 0.8963, precision 0.1876, recall 0.9405 17: African_crocodile: AP 0.8824, precision 0.1549, recall 0.9540 0: goldfish: AP 0.8808, precision 0.2061, recall 0.9072 9: European_fire_salamander: AP 0.8668, precision 0.7000, recall 0.8750 41: American_egret: AP 0.8539, precision 0.3143, recall 0.9167 30: jacamar: AP 0.8486, precision 0.3951, recall 0.8889 8: kite: AP 0.8458, precision 0.1894, recall 0.8776 61: jaguar: AP 0.8356, precision 0.4100, recall 0.9318 4: bulbul: AP 0.8249, precision 0.1502, recall 0.8636 63: tiger: AP 0.8243, precision 0.2222, recall 0.9412 10: axolotl: AP 0.8185, precision 0.2500, recall 0.8824 32: drake: AP 0.8087, precision 0.3529, recall 0.8372 58: Persian_cat: AP 0.8070, precision 0.2686, recall 0.9432 96: eel: AP 0.8002, precision 0.2476, recall 0.8636 94: indri: AP 0.7993, precision 0.1882, recall 0.8750 26: prairie_chicken: AP 0.7961, precision 0.3182, recall 0.8370 3: robin: AP 0.7944, precision 0.2390, recall 0.8444 28: macaw: AP 0.7895, precision 0.3564, recall 0.8182 99: puffer: AP 0.7706, precision 0.4375, recall 0.8000 86: chimpanzee: AP 0.7645, precision 0.3556, recall 0.7805 40: little_blue_heron: AP 0.7566, precision 0.1317, recall 0.8571 43: European_gallinule: AP 0.7503, precision 0.5424, recall 0.7805 6: chickadee: AP 0.7450, precision 0.3838, recall 0.8444 29: hornbill: AP 0.7387, precision 0.4337, recall 0.7826 39: flamingo: AP 0.7346, precision 0.5537, recall 0.7444 27: partridge: AP 0.7223, precision 0.5000, recall 0.7561 22: water_snake: AP 0.7150, precision 0.3333, recall 0.8140 65: brown_bear: AP 0.7100, precision 0.1635, recall 0.8293 42: crane: AP 0.6983, precision 0.3857, recall 0.7297 19: thunder_snake: AP 0.6908, precision 0.2547, recall 0.7941 47: killer_whale: AP 0.6897, precision 0.0785, recall 0.8293 14: common_iguana: AP 0.6868, precision 0.3699, recall 0.7941 16: Komodo_dragon: AP 0.6818, precision 0.2566, recall 0.7838 84: orangutan: AP 0.6251, precision 0.1236, recall 0.8462 72: porcupine: AP 0.6178, precision 0.2180, recall 0.7436 75: bison: AP 0.6174, precision 0.0678, recall 0.7667 34: goose: AP 0.6173, precision 0.2160, recall 0.7292 38: jellyfish: AP 0.6168, precision 0.1602, recall 0.7857 54: hyena: AP 0.6128, precision 0.2059, recall 0.8537 51: Newfoundland: AP 0.6114, precision 0.0434, recall 0.8049 90: marmoset: AP 0.6093, precision 0.2061, recall 0.7500 73: hippopotamus: AP 0.6078, precision 0.1133, recall 0.8718 48: Japanese_spaniel: AP 0.6058, precision 0.2589, recall 0.7250 45: pelican: AP 0.5936, precision 0.2542, recall 0.6818 70: tiger_beetle: AP 0.5911, precision 0.1081, recall 0.7273 1: great_white_shark: AP 0.5848, precision 0.4194, recall 0.6047 97: coho: AP 0.5843, precision 0.2381, recall 0.7500 46: king_penguin: AP 0.5773, precision 0.1263, recall 0.7705 62: lion: AP 0.5663, precision 0.4630, recall 0.6944 66: American_black_bear: AP 0.5610, precision 0.1538, recall 0.7500 98: sturgeon: AP 0.5587, precision 0.0982, recall 0.7755 85: gorilla: AP 0.5555, precision 0.1959, recall 0.7073 50: pug: AP 0.5507, precision 0.0730, recall 0.7021 33: red-breasted_merganser: AP 0.5406, precision 0.3529, recall 0.5714 52: Cardigan: AP 0.5350, precision 0.3919, recall 0.6591 71: hamster: AP 0.5348, precision 0.3333, recall 0.5714 93: Madagascar_cat: AP 0.5344, precision 0.1475, recall 0.7442 92: howler_monkey: AP 0.5330, precision 0.1954, recall 0.6939 69: meerkat: AP 0.5313, precision 0.1857, recall 0.7429 68: mongoose: AP 0.5236, precision 0.4237, recall 0.7143 55: kit_fox: AP 0.5195, precision 0.2972, recall 0.8182 64: cheetah: AP 0.5142, precision 0.1649, recall 0.7949 13: mud_turtle: AP 0.5063, precision 0.2095, recall 0.6875 81: black-footed_ferret: AP 0.5034, precision 0.1135, recall 0.8140 59: leopard: AP 0.4903, precision 0.2424, recall 0.6316 74: ox: AP 0.4841, precision 0.0929, recall 0.6176 79: weasel: AP 0.4821, precision 0.1605, recall 0.6047 15: Gila_monster: AP 0.4692, precision 0.2424, recall 0.6667 60: snow_leopard: AP 0.4615, precision 0.2857, recall 0.6829 77: impala: AP 0.4595, precision 0.2639, recall 0.5758 5: jay: AP 0.4550, precision 0.5405, recall 0.5000 24: diamondback: AP 0.4337, precision 0.2566, recall 0.7632 37: koala: AP 0.4210, precision 0.1709, recall 0.6250 53: white_wolf: AP 0.4198, precision 0.2468, recall 0.5429 76: ram: AP 0.4173, precision 0.0884, recall 0.5238 78: llama: AP 0.4078, precision 0.0682, recall 0.5854 67: ice_bear: AP 0.3905, precision 0.3134, recall 0.6364 57: tabby: AP 0.3879, precision 0.2500, recall 0.6829 95: African_elephant: AP 0.3853, precision 0.1316, recall 0.7447 36: wallaby: AP 0.3833, precision 0.1667, recall 0.5333 12: leatherback_turtle: AP 0.3825, precision 0.2472, recall 0.6667 88: colobus: AP 0.3751, precision 0.1031, recall 0.5952 35: echidna: AP 0.3738, precision 0.1981, recall 0.6000 25: trilobite: AP 0.3606, precision 0.2644, recall 0.6571 91: capuchin: AP 0.3591, precision 0.1605, recall 0.6667 89: proboscis_monkey: AP 0.3486, precision 0.2125, recall 0.4857 18: triceratops: AP 0.3427, precision 0.1556, recall 0.4516 20: ringneck_snake: AP 0.3333, precision 0.1073, recall 0.6111 21: hognose_snake: AP 0.2665, precision 0.2022, recall 0.4737 56: Arctic_fox: AP 0.2591, precision 0.1195, recall 0.6750 2: cock: AP 0.2361, precision 0.1102, recall 0.5000 87: siamang: AP 0.2317, precision 0.2277, recall 0.5750 49: Afghan_hound: AP 0.2289, precision 0.2154, recall 0.3415 23: boa_constrictor: AP 0.1518, precision 0.1477, recall 0.3611 80: mink: AP 0.1302, precision 0.1096, recall 0.4103 82: otter: AP 0.0959, precision 0.1705, recall 0.3191 mAP@IoU=0.50 result: 58.671837 mPrec@IoU=0.50 result: 24.441985 mRec@IoU=0.50 result: 72.986519 Evaluation time cost: 322.235986s

we can add this information to our presentation!