Time | Item | Who | Notes |
---|
| | Paul | - Using Hough transform to identify and remove dotted lines
- Accuracy is back around 70%
- All OSR cards have been completed for DESCENT fields
|
| | Sandeep | - Tried a few options to improve digit classification accuracy
- What remains to be done is to generate a model after training on the entire NIST digit dataset and see if that will make any improvement
- Otherwise concluding with improvements
- Will start running the extractor program from Bridges and start processing Type1OSR1890s cards
|
| | All | - Discussed Started discussions about the paper
- Will try to find a balance between domain specific and technical aspects of the project
- Will start putting things down in a Google doc that will be shared by Alison
|