[DEEP-180] Add mmsegmentation integration

[activesoull]

🚀 🚀 Pull Request

Impact

• Bug fix (non-breaking change which fixes expected existing functionality)
• Enhancement/New feature (adds functionality without impacting existing logic)
• Breaking change (fix or feature that would cause existing functionality to change)

Description

added segmentation integration to deeplake. As a reference segmentation version is set the v0.30.0

install steps

git clone https://github.com/open-mmlab/mmsegmentation.git 
git checkout v0.30.0
cd mmsegmentation
python3 -m pip install -e .

Specify the config file

seg_mask_config.py

rom mmdet.apis import set_random_seed

_base_ = '/home/ubuntu/mmsegmentation/configs/pspnet/pspnet_r101-d8_512x512_4x4_160k_coco-stuff164k.py'

load_from = '/home/ubuntu/mmsegmentation/checkpoints/pspnet_r50-d8_512x1024_40k_cityscapes_20200605_003338-2966598c.pth'

img_norm_cfg = dict(
    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
crop_size = (256, 256)

reduce_zero_label=False

train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations'),
    dict(type='Resize', img_scale=(320, 240), ratio_range=(0.5, 2.0)),
    dict(type='RandomCrop', crop_size=crop_size),
    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),
    dict(type='DefaultFormatBundle'),
    dict(type='Collect', keys=['img', 'gt_semantic_seg']),
    ]

test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(
        type='MultiScaleFlipAug',
        img_scale=(320, 240),
        # img_ratios=[0.5, 0.75, 1.0, 1.25, 1.5, 1.75],
        flip=False,
        transforms=[
            dict(type='Resize', img_scale=(320, 240), keep_ratio=True),
            dict(type='RandomFlip'),
            dict(type='Normalize', **img_norm_cfg),
            dict(type='ImageToTensor', keys=['img']),
            dict(type='Collect', keys=['img']),
        ])
]

evaluation = dict(metric=["mIoU"], interval=500)

data = dict(
    samples_per_gpu=4,
    workers_per_gpu=8,
    train=dict(
        pipeline=train_pipeline,
        deeplake_path="hub://activeloop/coco-train-seg-mask",
        deeplake_tensors = {"img": "images", "gt_semantic_seg": "seg_masks"},
        deeplake_dataloader={"shuffle": False, "num_workers": 0, "drop_last": True}
    ),
    val=dict(
        pipeline=test_pipeline,
        deeplake_path="hub://activeloop/coco-val-seg-mask"",
        deeplake_tensors = {"img": "images", "gt_semantic_seg": "seg_masks"},
        deeplake_dataloader={"shuffle": False, "batch_size": 1, "num_workers": 0, "drop_last": True}
    )
)

deeplake_dataloader_type = "c++"

work_dir = "./work_exps"

optimizer = dict(lr=0.02 / 8)
lr_config = dict(warmup=None)
log_config = dict(interval=50)
checkpoint_config = dict(interval=1000)


seed = None
gpu_ids = range(2)


runner = dict(type="IterBasedRunner", max_iters=1000, max_epochs=None)

device = "cuda"

in integration, we kept the openmlab's naming conventions that is why deeplake export two apis build_segmentor and train_segmentor

specify entry point

train.py

import os
import torch

from mmcv import Config
import mmcv
from deeplake.integrations import mmseg as mmseg_deeplake
import os

# from pytorch_memlab import LineProfiler


if __name__ == "__main__":
    current_loc = os.getcwd()
    cfg_file = f"{current_loc}/seg_mask_config.py"
    validate = True
    distributed = True

    # Read the config file
    cfg = Config.fromfile(cfg_file)
    cfg.model.decode_head.num_classes = 81
    cfg.model.auxiliary_head.num_classes = 81

    # build segmentor
    model = mmseg_deeplake.build_segmentor(
        cfg.model
    )

    # train_segmentor
    mmseg_deeplake.train_segmentor(
        model,
        cfg,
        distributed=distributed,
        validate=validate
    )

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 444 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
deeplake/integrations/mmseg/mmseg_.py	0.00%	310 Missing ⚠️
deeplake/integrations/mm/mm_common.py	0.00%	85 Missing ⚠️
deeplake/integrations/mmdet/mmdet_.py	0.00%	37 Missing ⚠️
deeplake/integrations/mmdet/mmdet_runners.py	0.00%	9 Missing ⚠️
deeplake/integrations/mmseg/__init__.py	0.00%	2 Missing ⚠️
deeplake/integrations/mmdet/mmdet_utils.py	0.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[sonarcloud]

Quality Gate failed

Failed conditions
6.2% Duplication on New Code (required ≤ 3%)

See analysis details on SonarCloud