why the eval result is not good?

[15241192958]

i finetuned 6000 own images with efficiendetd1 (in coco ckpt ), after 3hours ，the loss is 0.8,the AP50 is high ,but the accuracy is low 0.45,而且他会有很多误检(my Poor english)，what should i do to improve its AP?

[qtw1998]

i finetuned 6000 own images with efficiendetd1 (in coco ckpt ), after 3hours ，the loss is 0.8,the AP50 is high ,but the accuracy is low 0.45,而且他会有很多误检(my Poor english)，what should i do to improve its AP?

尝试改下anchor size

[elesun2018]

when i run eval code on coco with pretrained models, meet a problem as follows: automl0401/efficientdet# python main.py --mode=eval --model_name=${MODEL} --model_dir=${CKPT_PATH} --validation_file_pattern=tfrecord/val* --val_json_file=annotations/instances_val2017.json --hparams="use_bfloat16=false" --use_tpu=False

CKPT_PATH is the pretrained d0 model.

DONE (t=34.86s). Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.002 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.005 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.002 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.003 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.001 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.000 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.006 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.011 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.012 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.018 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.006 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.004 INFO:tensorflow:Finished evaluation at 2020-04-28-17:12:04 INFO:tensorflow:Finished evaluation at 2020-04-28-17:12:04 I0428 17:12:04.599111 140494972888896 evaluation.py:275] Finished evaluation at 2020-04-28-17:12:04 INFO:tensorflow:Saving dict for global step 0: AP = 0.0022378154, AP50 = 0.0052162167, AP75 = 0.0015809378, APl = 0.00015473206, APm = 0.0014778649, APs = 0.0029822278, ARl = 0.0036031725, ARm = 0.006258475, ARmax1 = 0.005900326, ARmax10 = 0.011014862, ARmax100 = 0.011780873, ARs = 0.018089531, box_loss = 0.0, cls_loss = 0.02997985, global_step = 0, loss = 0.11981841 I0428 17:12:04.600037 140494972888896 estimator.py:2049] Saving dict for global step 0: AP = 0.0022378154, AP50 = 0.0052162167, AP75 = 0.0015809378, APl = 0.00015473206, APm = 0.0014778649, APs = 0.0029822278, ARl = 0.0036031725, ARm = 0.006258475, ARmax1 = 0.005900326, ARmax10 = 0.011014862, ARmax100 = 0.011780873, ARs = 0.018089531, box_loss = 0.0, cls_loss = 0.02997985, global_step = 0, loss = 0.11981841 INFO:tensorflow:Saving 'checkpoint_path' summary for global step 0: models/pretrained_models/efficientdet-d1/model

such low AP values ? why ? thanks.

[zishanahmed08]

@elesun2018 Did u figure it out?