Project approach

  • Define task (Object detection, Colorization of line arts)
  • Collect dataset (MS Coco, Public web sites)
    • Search for academic datasets and baselines
    • Build your own (From Twitter, News, Website …)
  • Define the metrics
    • Search for established metrics
  • Clean and preprocess the data
    • Select features and transform data
      • One-hot vector, bag of words, bucketize
      • Logarithm scale, spectrogram
    • Remove noise or outliers
    • Remove invalid and duplicate data
    • Scaling or whiten data
  • Split datasets for training, validation and testing
    • Visualize and summarize dataset
    • Validate dataset
  • Establish a baseline
    • Implement a simple model
    • Compute metrics as a baseline
    • Analyze errors for area of improvements
  • Select network structure
    • CNN, LSTM…
  • Implement a deep network
    • Code debugging and validation
    • Parameter initialization
    • Compute metrics
    • Choose hyper-parameters
    • Visualize, validate and summarize result
    • Analyze errors
    • Refine
      • Add layers and nodes
      • Optimization tricks
  • Hyper-parameters fine tunings
  • Try other model variants

Deep learning resources

Listing of major papers: [https://github.com/kjw0612/awesome-deep-vision]
Machine learning competition: [https://www.kaggle.com/]

Research paper publication:

CVPR: IEEE Conference on Computer Vision and Pattern Recognition
ICCV: International Conference on Computer Vision
ECCV: European Conference on Computer Vision
NIPS: Neural Information Processing Systems
ICLR: International Conference on Learning Representations \

Dataset listing:

[http://www.cvpapers.com/datasets.html]
[http://riemenschneider.hayko.at/vision/dataset/]
[https://en.wikipedia.org/wiki/List_of_datasets_for_machine_learning_research]