xenon

Start with horizons at 30% from top, then 40% and then 50%
For each horizon, use box sizes from 0.25, 0.5, 1.0, 1.5 and 2.0 times trained BB sizes
Scan left to right with 0.2 of above box width
Catching kids: Kids will be a smaller box but on higher horizon. So a kid on 305 hoprizon will hopefully be identified by the smaller box at 50% horizon

Files:

xenon_CodeLab: Experimental code - To develop utility code and code tit-bits to be later used for building the assignment

Log:

08-Dec-2018: What does HoG do?

Can we take a simple 2D figure like a house shape and plot HoGs to see what it looks like?
How does it look for a rect.? A triangle? and then the house? A circle?
A steep line vs a gently sloped line

08-Dec-2018: Understanding what gradients mean for images

(Klette, 2014: 11): See 'Detecting Step-Edges by First- or Second-Order Derivatives'
1st order and 2nd order derivatives provide edge detection
The 'curve' on which to do derivatives are intensities? Along X or Y axes
Check using simple high contrast images developed in MS Paint
See if we can plot the intensities along sections of a square and triangle and see expected itensity peaks

REFERENCES:

Provide feedback

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
CreateData.py		CreateData.py
HoG_V2.py		HoG_V2.py
Intuition_building_Filters.ipynb		Intuition_building_Filters.ipynb
Intuition_building_Gradients.ipynb		Intuition_building_Gradients.ipynb
Intuition_building_HoG.ipynb		Intuition_building_HoG.ipynb
README.md		README.md
SplitInstancesIntoTrainingAndTestSets.py		SplitInstancesIntoTrainingAndTestSets.py
Task_01_HoG_Histogram_of_gradients.ipynb		Task_01_HoG_Histogram_of_gradients.ipynb
Task_02_Generate_training_instances.ipynb		Task_02_Generate_training_instances.ipynb
Task_03_Class_balanced_training_and_test_data.ipynb		Task_03_Class_balanced_training_and_test_data.ipynb
Task_04.a._RandomForests_model_HoG_features.ipynb		Task_04.a._RandomForests_model_HoG_features.ipynb
Task_04.b._RandomForests_model_PCA_features.ipynb		Task_04.b._RandomForests_model_PCA_features.ipynb
Task_05._RandomForests_hyperparameter_optimisation.ipynb		Task_05._RandomForests_hyperparameter_optimisation.ipynb
Task_06_SVM_model_HoG_features.ipynb		Task_06_SVM_model_HoG_features.ipynb
Task_07._SVM_hyperparameter_optimisation.ipynb		Task_07._SVM_hyperparameter_optimisation.ipynb
Task_08.b._Test_models_RF_optimised.ipynb		Task_08.b._Test_models_RF_optimised.ipynb
Task_08_Test_models_SVM_Optimised-CUSTOM_IMAGES.ipynb		Task_08_Test_models_SVM_Optimised-CUSTOM_IMAGES.ipynb
Task_08_Test_models_SVM_Optimised.ipynb		Task_08_Test_models_SVM_Optimised.ipynb
xenon_testing_10-Jan-2019.htm		xenon_testing_10-Jan-2019.htm