Jesus Arias Fisteus
Jesus Arias Fisteus
A user reported that captures where not saved into the "internal" directory in Windows. The session path contained accents. In a different session path with no accents, everything worked. The...
An ONNX model could be used: https://github.com/onnx/models/tree/main/vision/classification/mnist There are instructions there on the preprocessing needed. It can be loaded into OpenCV>=4.5 with: https://docs.opencv.org/4.x/d6/d0f/group__dnn.html#ga7faea56041d10c71dbbd6746ca854197 See also: https://docs.opencv.org/4.x/dc/d70/pytorch_cls_tutorial_dnn_conversion.html
On LaTeX errors print the LaTeX error formatted. Right now, the program shows unformatted text with literal "\n" values among other issues, which largely difficult reading error messages.
Update also the documentation for the first page of the new session wizard. Screenshots are available in the comments of issue #142.
The user activates this option in the menu but Eyegrade seems not to look for a new exam.
Support questions in which there may me more than one correct answer.
It would allow encoding more information than the infobits (the squares), opening the door to some improvements, like reading the exam id from the answer sheet. This is a long-term...
With maximum complexity set to 12: ``` eyegrade/qtgui/wizards.py:525:5: C901 'NewSessionPageScores._update_combo' is too complex (15) eyegrade/create_exam.py:243:1: C901 'create_exam' is too complex (36) eyegrade/detection.py:137:5: C901 'ExamDetector.detect' is too complex (22) eyegrade/detection.py:843:1: C901...
The call to `cv2.VideoCapture(camera_id)` blocks when the camera is busy. It returns once it has been released and can be acquired by the current call. The OpenCV API doesn't seem...
The `pdflatex` command converts EPS and PS images to PDF in order to include them in the LaTeX file. Similarly to the rest of the temporary files produced by `pdflatex`,...