Testing mIoU on PASCAL-Context got nan.

Hi,

I use the below data config.

dataset_type = 'PascalContextDataset'
data_root = '/home/ubuntu/dataset/PASCAL_Context/VOCdevkit/VOC2010/'
img_norm_cfg = dict(
    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
img_scale = (520, 520)
crop_size = (480, 480)
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations'),
    dict(type='Resize', img_scale=img_scale, ratio_range=(0.5, 2.0)),
    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
    dict(type='RandomFlip', flip_ratio=0.5),
    dict(type='PhotoMetricDistortion'),
    dict(type='Normalize', **img_norm_cfg),
    dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
    dict(type='DefaultFormatBundle'),
    dict(type='Collect', keys=['img', 'gt_semantic_seg']),
]
test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(
        type='MultiScaleFlipAug',
        img_scale=img_scale,
        img_ratios=[0.5, 0.75, 1.0, 1.25, 1.5, 1.75],
        flip=True,
        transforms=[
            dict(type='Resize', keep_ratio=True),
            dict(type='RandomFlip'),
            dict(type='Normalize', **img_norm_cfg),
            dict(type='ImageToTensor', keys=['img']),
            dict(type='Collect', keys=['img']),
        ])
]
data = dict(
    samples_per_gpu=1,
    workers_per_gpu=4,
    train=dict(
        type=dataset_type,
        data_root=data_root,
        img_dir='JPEGImages',
        ann_dir='SegmentationClass',
        split='ImageSets/Segmentation/train.txt',
        pipeline=train_pipeline),
    val=dict(
        type=dataset_type,
        data_root=data_root,
        img_dir='JPEGImages',
        ann_dir='SegmentationClass',
        split='ImageSets/Segmentation/val.txt',
        pipeline=test_pipeline),
    test=dict(
        type=dataset_type,
        data_root=data_root,
        img_dir='JPEGImages',
        ann_dir='SegmentationClass',
        split='ImageSets/Segmentation/val.txt',
        pipeline=test_pipeline))

The testing result is quite strange.

Class                  IoU        Acc
background           93.85      98.49
aeroplane            88.95      91.85
bicycle              63.43      81.58
bird                 88.93      92.60
boat                 69.97      72.82
bottle               75.46      89.79
bus                  84.96      87.64
car                  82.18      84.16
cat                  86.61      94.48
chair                24.36      30.57
cow                  87.53      93.38
table                46.55      50.18
dog                  80.99      90.12
horse                82.93      85.69
motorbike            82.89      91.69
person               83.98      91.55
pottedplant          51.72      57.72
sheep                81.93      83.15
sofa                 40.73      47.11
train                80.45      81.60
tvmonitor            66.43      72.46
bag                    nan        nan
bed                    nan        nan
bench                  nan        nan
book                   nan        nan
building               nan        nan
cabinet                nan        nan
ceiling                nan        nan
cloth                  nan        nan
computer               nan        nan
cup                    nan        nan
door                   nan        nan
fence                  nan        nan
floor                  nan        nan
flower                 nan        nan
food                   nan        nan
grass                  nan        nan
ground                 nan        nan
keyboard               nan        nan
light                  nan        nan
mountain               nan        nan
mouse                  nan        nan
curtain                nan        nan
platform               nan        nan
sign                   nan        nan
plate                  nan        nan
road                   nan        nan
rock                   nan        nan
shelves                nan        nan
sidewalk               nan        nan
sky                    nan        nan
snow                   nan        nan
bedclothes             nan        nan
track                  nan        nan
tree                   nan        nan
truck                  nan        nan
wall                   nan        nan
water                  nan        nan
window                 nan        nan
wood                   nan        nan
Summary:
Scope                 mIoU       mAcc       aAcc
global               73.56      79.46      94.18

Could you help me figure it out? Many thanks!

[yamengxi]

I guess you used the wrong parameter num_classes, please check if your network parameters are correct.

[ke-dev]

hi, @fangruizhu ,I have the same problem, has your problem been solved?