Dataset & Data preparetion
The database, the preparation of the same, the download of the model, and adding more layers to the end of the downloaded model (InceptionResNetV2) were done in the same way in the Transfer Learning & Classification of dogs and cats project.
Detection
For detection, an image is taken from the database or from another source, and it is sectioned into sub-images (parts of the original image) each sub-image is subjected to the InceptionResNetV2 model already trained and a box called bounding is generated. box to show which area the object goes into.
The detection in this project was proposed without the help of libraries, so it is not compared with the detectors in the state of the art.
More Results
Because the model was not taught to calcify the absence of a cat and dog, the model tends to classify regions where nothing of interest is found as a cat.
For more details, you can see the report here.
Annexes
Here, It can be seen another detection but focus on a different topic, the detection of crosswalks.