Cycle learning rate for unet

data/voc/train.txt

-319_4
-93_0
-491_1
-278_0
-536_3
-555_0
-184_0
-471_4
-510_0
-434_4
-304_1
-179_0
-341_1
-367_2
-322_2
-48_3
-285_0
-429_1
-493_3
-91_2
-219_1
-105_4
-140_4
-12_4
-398_0
-57_4
-534_1
-512_2
-186_2
-557_2
-343_3
-306_3
-320_0
-204_4
-118_1
-365_0
-241_4
-448_0
-529_4
-287_2
-29_2
-103_2
-8_0
-146_2
-160_1
-358_0
-125_1
-239_4
-97_4
-430_0
-77_1
-514_4
-32_1
-475_0
-180_4
-551_4
-14_2
-453_3
-416_3
-51_2
-202_2
-247_2
-488_0
-261_1
-138_4
-224_1
-509_1
-281_4
-383_3
-144_0
-339_1
-101_0
-127_3
-162_3
-477_2
-30_3
-75_3
-432_2
-53_0
-414_1
-451_1
-16_0
-245_0
-361_4
-200_0
-324_4
-226_3
-88_3
-263_3
-409_4
-381_1
-381_0
-88_2
-226_2
-263_2
-245_1
-159_4
-200_1
-414_0
-53_1
-530_4
-16_1
-451_0
-30_2
-477_3
-432_3
-75_2
-127_2
-162_2
-258_4
-144_1
-101_1
-339_0
-383_2
-509_0
-468_4
-261_0
-488_1
-345_4
-224_0
-300_4
-202_3
-247_3
-453_2
-14_3
-51_3
-416_2
-77_0
-430_1
-475_1
-32_0
-358_1
-495_4
-160_0
-125_0
-8_1
-103_3
-146_3
-29_3
-287_3
-448_1
-385_4
-118_0
-320_1
-365_1
-343_2
-306_2
-512_3
-557_3
-186_3
-398_1
-455_4
-534_0
-410_4
-219_0
-378_4
-493_2
-91_3
-548_4
-199_4
-285_1
-429_0
-48_2
-367_3
-322_3
-304_0
-220_4
-341_0
-179_1
-265_4
-36_4
-184_1
-298_4
-555_1
-73_4
-510_1
-536_2
-121_4
-93_1
-164_4
-278_1
-491_0
-491_2
-278_3
-93_3
-510_3
-555_3
-184_3
-457_4
-536_0
-412_4
-322_1
-367_1
-179_3
-341_2
-304_2
-429_2
-285_3
-387_4
-48_0
-219_2
-123_4
-91_1
-166_4
-493_0
-34_4
-186_1
-557_1
-71_4
-512_1
-534_2
-398_3
-365_3
-320_3
-118_2
-306_0
-222_4
-343_0
-267_4
-287_1
-29_1
-448_3
-125_2
-160_2
-358_3
-146_1
-103_1
-8_3
-416_0
-51_1
-532_4
-14_1
-453_0
-32_2
-475_3
-430_3
-77_2
-224_2
-488_3
-261_2
-247_1
-202_1
-383_0
-509_2
-497_4
-162_0
-127_0
-339_2
-101_3
-144_3
-451_2
-16_3
-53_3
-414_2
-75_0
-432_1
-477_1
-30_0
-263_0
-347_4
-226_0
-88_0
-302_4
-200_3
-245_3
-381_2
-68_4
-283_4
-381_3
-200_2
-245_2
-263_1
-88_1
-226_1
-432_0
-75_1
-516_4
-30_1
-477_0
-182_4
-553_4
-16_2
-451_3
-414_3
-53_2
-101_2
-339_3
-144_2
-162_1
-127_1
-95_4
-509_3
-383_1
-247_0
-363_4
-202_0
-326_4
-224_3
-261_3
-488_2
-475_2
-32_3
-77_3
-430_2
-51_0
-416_1
-453_1
-14_0
-146_0
-8_2
-103_0
-125_3
-358_2
-160_3
-448_2
-29_0
-287_0
-306_1
-343_1
-365_2
-118_3
-320_2
-534_3
-398_2
-557_0
-186_0
-473_4
-512_0
-436_4
-91_0
-493_1
-219_3
-48_1
-429_3
-285_2
-341_3
-179_2
-304_3
-322_0
-206_4
-367_0
-243_4
-10_4
-55_4
-536_1
-510_2
-184_2
-555_2
-278_2
-491_3
-93_2
-107_4
-142_4
-298_3
-36_3
-471_2
-434_2
-73_3
-412_1
-55_0
-10_0
-457_1
-142_0
-107_0
-319_2
-121_3
-164_3
-199_3
-548_3
-387_1
-243_0
-367_4
-206_0
-322_4
-220_3
-265_3
-71_1
-436_0
-512_4
-473_0
-34_1
-186_4
-557_4
-455_3
-12_2
-57_2
-410_3
-105_2
-140_2
-378_3
-166_1
-123_1
-91_4
-287_4
-29_4
-385_3
-529_2
-204_2
-241_2
-267_1
-222_1
-14_4
-51_4
-532_1
-514_2
-180_2
-551_2
-495_3
-239_2
-97_2
-103_4
-146_4
-281_2
-468_3
-345_3
-300_3
-138_2
-326_0
-202_4
-363_0
-247_4
-530_3
-553_0
-182_0
-477_4
-516_0
-432_4
-95_0
-497_1
-258_3
-409_2
-283_0
-68_0
-302_1
-347_1
-159_3
-361_2
-324_2
-361_3
-159_2
-324_3
-302_0
-88_4
-226_4
-347_0
-263_4
-283_1
-68_1
-409_3
-258_2
-127_4
-95_1
-162_4
-497_0
-30_4
-182_1
-553_1
-75_4
-516_1
-530_2
-326_1
-363_1
-345_2
-138_3
-300_2
-468_2
-281_3
-383_4
-495_2
-97_3
-239_3
-514_3
-551_3
-180_3
-453_4
-532_0
-416_4
-267_0
-343_4
-222_0
-306_4
-204_3
-241_3
-385_2
-529_3
-493_4
-166_0
-123_0
-105_3
-378_2
-140_3
-12_3
-455_2
-410_2
-57_3
-436_1
-71_0
-34_0
-473_1
-220_2
-265_2
-243_1
-206_1
-48_4
-387_0
-548_2
-199_2
-121_2
-319_3
-164_2
-142_1
-107_1
-55_1
-412_0
-536_4
-457_0
-10_1
-471_3
-36_2
-298_2
-73_2
-434_3
-10_3
-457_2
-412_2
-55_3
-434_1
-73_0
-298_0
-36_0
-471_1
-491_4
-164_0
-319_1
-121_0
-107_3
-142_3
-387_2
-429_4
-199_0
-548_0
-265_0
-341_4
-220_0
-304_4
-206_3
-243_3
-57_1
-410_0
-534_4
-455_0
-12_1
-473_3
-34_2
-71_2
-436_3
-123_2
-166_2
-140_1
-378_0
-219_4
-105_1
-529_1
-385_0
-222_2
-267_2
-241_1
-204_1
-118_4
-32_4
-180_1
-551_1
-77_4
-514_1
-532_2
-125_4
-239_1
-97_1
-160_4
-495_0
-281_1
-468_0
-509_4
-363_3
-326_3
-300_0
-138_1
-224_4
-345_0
-261_4
-516_3
-553_3
-182_3
-451_4
-530_0
-414_4
-339_4
-258_0
-497_2
-95_3
-68_3
-283_3
-381_4
-409_1
-324_1
-159_0
-361_1
-347_2
-302_2
-347_3
-302_3
-324_0
-200_4
-361_0
-245_4
-159_1
-409_0
-68_2
-283_2
-497_3
-95_2
-101_4
-258_1
-144_4
-16_4
-53_4
-530_1
-516_2
-182_2
-553_2
-138_0
-300_1
-488_4
-345_1
-363_2
-326_2
-468_1
-281_0
-97_0
-239_0
-358_4
-495_1
-8_4
-532_3
-551_0
-180_0
-475_4
-514_0
-430_4
-241_0
-365_4
-204_0
-320_4
-222_3
-267_3
-529_0
-448_4
-385_1
-378_1
-140_0
-105_0
-123_3
-166_3
-34_3
-473_2
-436_2
-71_3
-410_1
-57_0
-398_4
-12_0
-455_1
-206_2
-243_2
-179_4
-265_1
-220_1
-548_1
-199_1
-285_4
-387_3
-107_2
-142_2
-164_1
-278_4
-121_1
-93_4
-319_0
-73_1
-434_0
-510_4
-471_0
-36_1
-184_4
-298_1
-555_4
-457_3
-10_2
-55_2
-412_3
-126_0
-163_0
-496_4
-145_3
-338_2
-100_3
-415_2
-52_3
-17_3
-450_2
-31_0
-476_1
-433_1
-74_0
-89_0
-227_0
-303_4
-262_0
-346_4
-244_3
-201_3
-380_2
-161_2
-359_3
-124_2
-102_1
-9_3
-147_1
-452_0
-15_1
-50_1
-417_0
-533_4
-76_2
-431_3
-474_3
-33_2
-260_2
-489_3
-225_2
-203_1
-246_1
-382_0
-508_2
-218_2
-492_0
-167_4
-90_1
-122_4
-70_4
-513_1
-187_1
-35_4
-556_1
-399_3
-535_2
-321_3
-119_2
-364_3
-342_0
-266_4
-307_0
-223_4
-28_1
-286_1
-449_3
-92_3
-279_3
-490_2
-554_3
-185_3
-511_3
-537_0
-413_4
-456_4
-366_1
-323_1
-305_2
-178_3
-340_2
-284_3
-428_2
-49_0
-386_4
-49_1
-284_2
-428_3
-305_3
-340_3
-178_2
-366_0
-242_4
-323_0
-207_4
-54_4
-537_1
-11_4
-185_2
-554_2
-511_2
-92_2
-490_3
-279_2
-143_4
-106_4
-449_2
-286_0
-28_0
-342_1
-307_1
-119_3
-321_2
-364_2
-399_2
-535_3
-513_0
-437_4
-556_0
-472_4
-187_0
-492_1
-90_0
-218_3
-508_3
-382_1
-203_0
-327_4
-246_0
-362_4
-489_2
-260_3
-225_3
-431_2
-76_3
-33_3
-474_2
-15_0
-452_1
-417_1
-50_0
-9_2
-102_0
-147_0
-359_2
-161_3
-124_3
-282_4
-69_4
-380_3
-244_2
-201_2
-227_1
-89_1
-262_1
-183_4
-476_0
-31_1
-552_4
-74_1
-433_0
-517_4
-52_2
-415_3
-450_3
-17_2
-145_2
-100_2
-338_3
-94_4
-126_1
-163_1
-338_1
-100_0
-145_0
-163_3
-126_3
-433_2
-74_3
-31_3
-476_2
-17_0
-450_1
-415_1
-52_0
-201_0
-325_4
-244_0
-360_4
-262_3
-89_3
-227_3
-380_1
-408_4
-147_2
-102_2
-9_0
-96_4
-124_1
-238_4
-161_1
-359_0
-181_4
-474_0
-33_1
-550_4
-76_1
-431_0
-515_4
-50_2
-417_3
-452_3
-15_2
-246_2
-203_2
-139_4
-225_1
-260_1
-489_0
-280_4
-508_1
-382_3
-90_2
-492_3
-141_4
-218_1
-104_4
-56_4
-535_1
-13_4
-399_0
-187_2
-556_2
-513_2
-307_3
-342_3
-364_0
-240_4
-321_0
-205_4
-119_1
-528_4
-449_0
-28_2
-286_2
-279_0
-490_1
-318_4
-92_0
-537_3
-511_0
-435_4
-554_0
-470_4
-185_0
-178_0
-340_1
-305_1
-323_2
-366_2
-49_3
-428_1
-284_0
-428_0
-549_4
-198_4
-284_1
-49_2
-323_3
-366_3
-340_0
-178_1
-264_4
-305_0
-221_4
-72_4
-511_1
-185_1
-299_4
-37_4
-554_1
-537_2
-490_0
-165_4
-279_1
-92_1
-120_4
-286_3
-28_3
-449_1
-384_4
-364_1
-119_0
-321_1
-307_2
-342_2
-556_3
-187_3
-513_3
-535_0
-411_4
-399_1
-454_4
-379_4
-218_0
-90_3
-492_2
-382_2
-469_4
-508_0
-225_0
-301_4
-489_1
-260_0
-344_4
-246_3
-203_3
-417_2
-50_3
-15_3
-452_2
-33_0
-474_1
-431_1
-76_0
-124_0
-359_1
-161_0
-494_4
-147_3
-9_1
-102_3
-380_0
-262_2
-227_2
-89_2
-201_1
-244_1
-158_4
-450_0
-17_1
-52_1
-415_0
-531_4
-74_2
-433_3
-476_3
-31_2
-163_2
-126_2
-100_1
-338_0
-259_4
-145_1
-552_3
-183_3
-517_3
-531_0
-415_4
-450_4
-259_0
-338_4
-94_3
-496_2
-282_3
-69_3
-408_1
-380_4
-158_0
-360_1
-325_1
-303_2
-346_2
-76_4
-515_1
-181_1
-33_4
-550_1
-533_2
-494_0
-161_4
-96_1
-124_4
-238_1
-508_4
-469_0
-280_1
-327_3
-362_3
-344_0
-260_4
-301_0
-139_1
-225_4
-13_1
-454_0
-411_0
-56_1
-535_4
-437_3
-70_2
-35_2
-472_3
-167_2
-122_2
-218_4
-104_1
-141_1
-379_0
-384_0
-528_1
-266_2
-223_2
-205_1
-119_4
-240_1
-54_3
-413_2
-456_2
-11_3
-470_1
-37_0
-299_0
-72_0
-435_1
-318_1
-120_0
-165_0
-490_4
-143_3
-106_3
-386_2
-198_0
-549_0
-428_4
-221_0
-305_4
-264_0
-340_4
-242_3
-207_3
-242_2
-207_2
-221_1
-178_4
-264_1
-549_1
-198_1
-284_4
-386_3
-143_2
-106_2
-92_4
-120_1
-318_0
-165_1
-279_4
-185_4
-299_1
-37_1
-470_0
-554_4
-435_0
-72_1
-511_4
-413_3
-54_2
-11_2
-456_3
-205_0
-321_4
-240_0
-364_4
-266_3
-223_3
-449_4
-384_1
-528_0
-104_0
-379_1
-141_0
-167_3
-122_3
-70_3
-437_2
-472_2
-35_3
-399_4
-454_1
-13_0
-56_0
-411_1
-489_4
-344_1
-139_0
-301_1
-327_2
-362_2
-280_0
-469_1
-359_4
-494_1
-238_0
-96_0
-9_4
-533_3
-515_0
-431_4
-550_0
-474_4
-181_0
-303_3
-346_3
-360_0
-244_4
-158_1
-325_0
-201_4
-408_0
-282_2
-69_2
-94_2
-496_3
-259_1
-145_4
-100_4
-52_4
-531_1
-17_4
-183_2
-552_2
-517_2
-531_3
-517_0
-433_4
-552_0
-476_4
-183_0
-496_1
-94_0
-259_3
-408_2
-69_0
-282_0
-346_1
-303_1
-325_2
-158_3
-360_2
-50_4
-533_1
-15_4
-181_2
-550_2
-515_2
-96_2
-238_2
-494_3
-147_4
-102_4
-469_3
-280_2
-301_3
-139_2
-344_3
-362_0
-246_4
-327_0
-203_4
-187_4
-35_1
-472_0
-556_4
-437_0
-70_1
-513_4
-411_3
-56_2
-13_2
-454_3
-141_2
-379_3
-104_2
-90_4
-122_1
-167_1
-28_4
-286_4
-528_2
-384_3
-240_2
-205_2
-223_1
-266_1
-72_3
-435_2
-470_2
-37_3
-299_3
-456_1
-11_0
-54_0
-413_1
-106_0
-143_0
-165_3
-318_2
-120_3
-198_3
-549_3
-386_1
-207_0
-323_4
-242_0
-366_4
-264_3
-221_3
-264_2
-221_2
-207_1
-242_1
-386_0
-49_4
-549_2
-198_2
-165_2
-120_2
-318_3
-106_1
-143_1
-11_1
-456_0
-413_0
-54_1
-537_4
-435_3
-72_2
-299_2
-37_2
-470_3
-223_0
-307_4
-266_0
-342_4
-240_3
-205_3
-528_3
-384_2
-122_0
-167_0
-492_4
-379_2
-141_3
-104_3
-56_3
-411_2
-454_2
-13_3
-472_1
-35_0
-70_0
-437_1
-362_1
-327_1
-139_3
-301_2
-344_2
-280_3
-469_2
-382_4
-238_3
-96_3
-494_2
-550_3
-181_3
-515_3
-533_0
-417_4
-452_4
-325_3
-360_3
-158_2
-346_0
-262_4
-303_0
-227_4
-89_4
-69_1
-282_1
-408_3
-259_2
-496_0
-163_4
-94_1
-126_4
-74_4
-517_1
-183_1
-31_4
-552_1
-531_2
-41_0
-406_1
-443_1
-465_2
-22_3
-67_3
-420_2
-135_3
-348_2
-170_3
-156_0
-113_0
-393_1
-519_3
-234_3
-271_3
-498_2
-257_0
-373_4
-212_0
-336_4
-441_3
-404_3
-43_2
-422_0
-65_1
-506_4
-20_1
-467_0
-192_4
-543_4
-172_1
-137_1
-85_4
-111_2
-329_3
-154_2
-391_3
-78_4
-293_4
-273_1
-98_1
-236_1
-210_2
-255_2
-500_2
-194_2
-545_2
-563_1
-45_4
-526_1
-117_4
-152_4
-268_2
-481_3
-83_2
-439_3
-295_2
-58_1
-332_0
-1_3
-216_4
-377_0
-253_4
-351_3
-169_2
-314_3
-547_0
-196_0
-463_4
-502_0
-426_4
-524_3
-388_2
-561_3
-209_3
-81_0
-483_1
-39_0
-297_0
-458_2
-375_2
-108_3
-3_1
-330_2
-316_1
-353_1
-316_0
-232_4
-353_0
-277_4
-375_3
-330_3
-3_0
-108_2
-458_3
-297_1
-39_1
-133_4
-81_1
-176_4
-483_0
-209_2
-524_2
-561_2
-388_3
-24_4
-196_1
-547_1
-61_4
-502_1
-169_3
-351_2
-314_2
-1_2
-332_1
-377_1
-397_4
-58_0
-439_2
-295_3
-481_2
-268_3
-83_3
-563_0
-447_4
-526_0
-402_4
-500_3
-545_3
-194_3
-210_3
-7_4
-255_3
-273_0
-357_4
-236_0
-98_0
-312_4
-391_2
-329_2
-111_3
-154_3
-487_4
-172_0
-137_0
-65_0
-422_1
-467_1
-20_0
-441_2
-43_3
-404_2
-257_1
-212_1
-234_2
-498_3
-271_2
-519_2
-393_0
-19_4
-156_1
-113_1
-135_2
-170_2
-348_3
-22_2
-465_3
-420_3
-67_2
-406_0
-41_1
-522_4
-443_0
-67_0
-420_1
-465_1
-22_0
-443_2
-41_3
-406_2
-113_3
-156_3
-348_1
-485_4
-170_0
-135_0
-519_0
-478_4
-393_2
-212_3
-5_4
-257_3
-271_0
-498_1
-355_4
-234_0
-310_4
-20_2
-467_3
-422_3
-65_2
-404_0
-43_1
-520_4
-441_0
-154_1
-248_4
-111_1
-329_0
-137_2
-172_2
-391_0
-149_4
-255_1
-210_1
-98_2
-236_2
-273_2
-526_2
-563_2
-26_4
-288_4
-194_1
-545_1
-63_4
-500_1
-131_4
-83_1
-268_1
-174_4
-481_0
-58_2
-558_4
-295_1
-189_4
-439_0
-314_0
-230_4
-351_0
-275_4
-169_1
-377_3
-332_3
-1_0
-388_1
-561_0
-445_4
-524_0
-400_4
-502_3
-547_3
-196_3
-483_2
-81_3
-209_0
-368_4
-458_1
-395_4
-39_3
-297_3
-353_2
-316_2
-108_0
-3_2
-330_1
-375_1
-330_0
-3_3
-108_1
-214_4
-375_0
-251_4
-353_3
-316_3
-297_2
-39_2
-458_0
-539_4
-115_4
-209_1
-150_4
-483_3
-81_2
-502_2
-196_2
-547_2
-561_1
-388_0
-47_4
-524_1
-377_2
-1_1
-332_2
-314_1
-169_0
-351_1
-295_0
-439_1
-58_3
-309_4
-83_0
-481_1
-268_0
-545_0
-194_0
-461_4
-500_0
-424_4
-526_3
-563_3
-236_3
-98_3
-273_3
-255_0
-371_4
-210_0
-334_4
-419_4
-391_1
-137_3
-172_3
-154_0
-329_1
-111_0
-43_0
-404_1
-441_1
-467_2
-20_3
-65_3
-422_2
-498_0
-271_1
-234_1
-128_4
-212_2
-257_2
-393_3
-519_1
-291_4
-170_1
-348_0
-229_4
-135_1
-87_4
-113_2
-156_2
-443_3
-406_3
-41_2
-420_0
\ No newline at end of file
+561
+97
+101
+357
+81
+220
+69
+55
+179
+439
+573
+132
+388
+183
+523
+113
+133
+448
+170
+157
+91
+64
+421
+161
+440
+233
+160
+309
+87
+118
+462
+67
+458
+226
+405
+82
+442
+115
+504
+394
+27
+300
+581
+272
+143
+537
+150
+520
+302
+147
+10
+417
+308
+553
+95
+196
+155
+127
+434
+297
+505
+531
+2
+154
+490
+345
+60
+254
+578
+478
+23
+386
+180
+73
+65
+538
+390
+444
+529
+231
+251
+568
+369
+203
+469
+124
+271
+76
+208
+86
+33
+159
+517
+471
+298
+455
+565
+404
+411
+166
+140
+371
+240
+327
+428
+243
+270
+215
+274
+306
+549
+255
+413
+387
+562
+445
+449
+299
+545
+38
+72
+149
+214
+8
+198
+206
+397
+427
+195
+74
+6
+32
+495
+416
+262
+263
+273
+524
+144
+525
+474
+44
+365
+88
+266
+83
+389
+264
+319
+156
+563
+205
+360
+39
+227
+572
+446
+401
+486
+232
+536
+383
+265
+245
+201
+287
+314
+119
+443
+547
+552
+92
+362
+165
+4
+433
+85
+367
+250
+521
+435
+503
+185
+221
+96
+481
+292
+368
+325
+15
+175
+564
+311
+138
+189
+289
+406
+194
+589
+588
+30
+475
+502
+102
+355
+468
+269
+373
+258
+137
+9
+396
+89
+146
+318
+480
+337
+493
+376
+224
+294
+117
+141
+548
+98
+202
+483
+148
+575
+459
+162
+234
+317
+130
+482
+510
+560
+108
+380
+400
+466
+583
+230
+541
+296
+582
+19
+22
+286
+126
+122
+290
+42
+377
+408
+399
+43
+431
+169
+403
+382
+544
+242
+249
+358
+24
+152
+425
+378
+282
+479
+379
+293
+341
+239
+375
+410
+51
+315
+123
+257
+191
+513
+307
+71
+129
+188
+577
+7
+181
+241
+361
+291
+142
+285
+514
+450
+328
+80
+465
+63
+5
+585
+158
+571
+225
+441
+53
+268
+267
+366
+320
+14
+500
+99
+447
+527
+496
+463
+301
+284
+45
+49
+567
+494
+453
+178
+586
+420
+332
+322
+432
+492
+173
+182
+100
+106
+556
+37
+28
+346
+558
+107
+574
+551
+550
+75
+204
+17
+370
+507
+451
+436
+94
+580
+385
+103
+57
+528
+279
+509
+66
+519
+535
+350
+488
+93
+168
+363
+247
+77
+554
+331
+526
+508
+90
+429
+120
+515
+342
+518
+18
+334
+110
+460
+276
+200
+151
+344
+236
+312
+246
+339
+177
+477
+392
+534
+384
+506
+415
+21
+235
+164
+414
+391
+356
+330
+217
+222
+470
+393
+349
+277
+25
+559
+228
+34
+219
+259
+423
+398
+192
+430
+210
+473
+176
+40
+58
+1
+354
+193
+41
+199
+539
+121
+238
+438
+374
+125
+338
+484
+348
+253
+305
+135
+569
+244
+498
+213
+261
+280
+186
+329
+61
+111
+229
+131
+184
+260
+584
+464
+372
+333
+487
+591
+79
+145
+216
+295
+461
+223
+59
+275
+530
+454
+456
+35
+424
+153
+20
+256
+418
+522
+283
+52
+190
+499
+62
+491
+278
+511
+587
+546
+343
+335
+172
+174
+304
+457
+187

mrnet/mrnet_module.py

@@ -42,7 +42,8 @@ class MultiUnet(nn.Module):
         self.pool = nn.MaxPool2d(2)
         self.outconv = nn.Sequential(
             nn.Conv2d(self.res9.outc, n_classes, kernel_size = 1),
-            nn.Sigmoid()
+            nn.Softmax()
+            #nn.Sigmoid()
         )
         # self.outconv = nn.Conv2d(self.res9.outc, n_classes,kernel_size = 1)
 

mrnet/train.py

@@ -28,8 +28,8 @@ def train_net(net, device, epochs = 5, batch_size = 1, lr = 0.1):
     val_loader = DataLoader(evalset, batch_size = batch_size, shuffle = False, num_workers = 8, pin_memory = True)
 
     optimizer = optim.Adam(net.parameters(), lr = lr)
-    criterion = nn.BCELoss()#nn.BCEWithLogitsLoss()
-    scheduler = lr_scheduler.StepLR(optimizer,30,0.5)#lr_scheduler.ReduceLROnPlateau(optimizer, 'min')
+    criterion = nn.BCELoss()# nn.BCEWithLogitsLoss()
+    scheduler = lr_scheduler.StepLR(optimizer, 30, 0.5)# lr_scheduler.ReduceLROnPlateau(optimizer, 'min')
 
     for epoch in range(epochs):
         net.train()

train_ignite.py

@@ -7,7 +7,7 @@ from torch.utils.data import DataLoader
 from torch.optim import lr_scheduler
 from ignite.contrib.handlers.param_scheduler import LRScheduler
 from ignite.engine import Events, create_supervised_trainer, create_supervised_evaluator
-from ignite.metrics import Accuracy, Loss, DiceCoefficient, ConfusionMatrix, RunningAverage
+from ignite.metrics import Accuracy, Loss, DiceCoefficient, ConfusionMatrix, RunningAverage,mIoU
 from ignite.contrib.handlers import ProgressBar
 from argparse import ArgumentParser
 
@@ -34,11 +34,13 @@ def run(train_batch_size, val_batch_size, epochs, lr):
     optimizer = optim.Adam(model.parameters(), lr = lr)
     cm = ConfusionMatrix(num_classes = 1)
     dice = DiceCoefficient(cm)
+    iou = mIoU(cm)
     loss = torch.nn.BCELoss()  # torch.nn.NLLLoss()
-    scheduler = LRScheduler(lr_scheduler.ReduceLROnPlateau(optimizer))
+    scheduler = LRScheduler(lr_scheduler.StepLR(optimizer, 30, 0.5))
 
     trainer = create_supervised_trainer(model, optimizer, loss, device = device)
-    evaluator = create_supervised_evaluator(model, metrics = {'accuracy': Accuracy(), 'dice': dice, 'nll': Loss(loss)},
+    evaluator = create_supervised_evaluator(model,
+                                            metrics = {'accuracy': Accuracy(), 'dice': dice, 'nll': Loss(loss)},
                                             device = device)
     RunningAverage(output_transform = lambda x: x).attach(trainer, 'loss')
     trainer.add_event_handler(Events.EPOCH_COMPLETED, scheduler)

unet/train.py

@@ -47,8 +47,8 @@ def train_net(net, device, epochs = 5, batch_size = 1, lr = 0.1, save_cp = True)
 
     # optimizer = optim.Adam(net.parameters(), lr=lr, weight_decay = 1e-8)
     optimizer = optim.RMSprop(net.parameters(), lr = lr, weight_decay = 1e-8)
-    scheduler = lr_scheduler.ReduceLROnPlateau(optimizer, 'min')
-    # criterion = nn.BCEWithLogitsLoss()
+    #scheduler = lr_scheduler.ReduceLROnPlateau(optimizer, 'min')
+    scheduler = lr_scheduler.CyclicLR(optimizer, base_lr = 1e-10, max_lr = 0.01)
     if net.n_classes > 1:
         criterion = nn.CrossEntropyLoss()
     else:
@@ -59,13 +59,6 @@ def train_net(net, device, epochs = 5, batch_size = 1, lr = 0.1, save_cp = True)
         epoch_loss = 0
         with tqdm(total = n_train, desc = f'Epoch {epoch + 1}/{epochs}', unit = 'img') as pbar:
             for imgs,true_masks in train_loader:
-                # imgs = batch['image']
-                # true_masks = batch['mask']
-                # assert imgs.shape[1] == net.n_channels, \
-                #     f'Network has been defined with {net.n_channels} input channels, ' \
-                #     f'but loaded images have {imgs.shape[1]} channels. Please check that ' \
-                #     'the images are loaded correctly.'
-
                 imgs = imgs.to(device = device, dtype = torch.float32)
                 mask_type = torch.float32 if net.n_classes == 1 else torch.long
                 true_masks = true_masks.to(device = device, dtype = mask_type)
@@ -80,11 +73,11 @@ def train_net(net, device, epochs = 5, batch_size = 1, lr = 0.1, save_cp = True)
                 optimizer.zero_grad()
                 loss.backward()
                 optimizer.step()
-
+                scheduler.step()
                 pbar.update(imgs.shape[0])
                 global_step += 1  # if global_step % (len(dataset) // (10 * batch_size)) == 0:
         val_score = eval_net(net, val_loader, device, n_val)
-        scheduler.step(val_score)
+        #scheduler.step(val_score)
         if net.n_classes > 1:
             logging.info('Validation cross entropy: {}'.format(val_score))
             writer.add_scalar('Loss/test', val_score, global_step)

utils/eval.py

@@ -2,6 +2,7 @@ import torch
 import torch.nn.functional as F
 from tqdm import tqdm
 from sklearn.metrics import jaccard_score
+import numpy as np
 
 from utils.dice_loss import dice_coeff, dice_coef
 
@@ -42,6 +43,7 @@ def eval_jac(net, loader, device, n_val):
 
             pred_masks = torch.round(pred_masks).cpu().detach().numpy()
             true_masks = torch.round(true_masks).cpu().numpy()
+            pred_masks = np.array([1 if x>0 else 0 for x in pred_masks])
             jac += jaccard_score(true_masks.flatten(), pred_masks.flatten())
 
             pbar.update(imgs.shape[0])