I. Quick-start on VNC (based on example-annos)

1) Open the existed Anaconda-env:

source activate open-mmlab

2.1). Train from the pre-saved images and annotations in the path ./mmdetection/data/CocoCust:

├── mmdetection
    |── data
        ├── CocoCust
            └── annotations
            └── coco
            └── fakeKitti

python tools/train.py configs/fcos/fcos_r50_caffe_fpn_gn-head_1x_coco_cust.py --gpus 1 --work-dir ./fcos_r50_fpn_gn_coco_cust

2.2). Train from the pre-saved annotations and load images from other paths:

├── mmdetection
    |── data
        ├── CocoCust
            └── annotations

python tools/train.py configs/fcos/fcos_r50_caffe_fpn_gn-head_1x_coco_cust_loadFromSeparatedFile.py --gpus 1 --work-dir ./fcos_r50_fpn_gn_coco_cust

Here, the example-data already exists in this repo.

Note 1.: As an example, train2017 is same to val2017 which will indicate how the model overfit our training-data.

        2.: Here, 'workers_per_gpu' was set to 0. If more workers were set, if will occurs an out of shared memory ERROR due to a docker environment configuretion.

3) Visualization in Tensorboard:

tensorboard --logdir=./fcos_r50_fpn_gn_coco_cust/tf_logs/ --bind_all

4) Visualize the training progress:

http://127.0.0.1:6006/

II. Quick-start on Pinky (based on example-annos)

1) Attch to the existed docker container:

docker attach mmdetection

2) Other steps are same as on VNC:

Note: As an example, train2017 is same to val2017 which will indicate how the model overfit our training-data.

III.Below are mmdetection-toolbox Installation from official repo.

Installation

Please refer to get_started.md for installation.

Getting Started

Please see get_started.md for the basic usage of MMDetection. We provide colab tutorial, and full guidance for quick run with existing dataset and with new dataset for beginners. There are also tutorials for finetuning models, adding new dataset, designing data pipeline, customizing models, customizing runtime settings and useful tools.

Please refer to FAQ for frequently asked questions.

Acknowledgement

MMClassification is an open source project that is contributed by researchers and engineers from various colleges and companies. We appreciate all the contributors who implement their methods or add new features, as well as users who give valuable feedbacks. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new classifiers.