intro

resource for store initial videos extract for extract frames in videos result for store images and databases obtained in tracings

method1: steady frames analysis

1 corner detection

code_extraction:

data_extract.ipynb to extract frames and store in extract files

extract_fourmis.ipynb to locate each ant (metrics of corner detection also in it)

extract_fourmis1.ipynb to locate ants massively

cnn count:

recognition_img_fix: cnn_generate.ipynb to create train/test set according to 14 imgs

Generate.ipynb rotate images for more training info

cnn_train to define the cnn model, and stock the model im model_frac.pkl (metrics of cnn also in it)

Preprocessing parameters stocked in mu.npy and sig.npy

test_mass.ipynb to see whether the pre-trained model works

nid_count.py to output the result in video

method2: differential to find moving objects

recognition_opencv;

open_cv.py to find moving ants and store their ids and locations in result in csv Track.py simple algo1 real time dict

merics in analyse.ipynb

find_trace: distinct.ipynb a try to combine the traces, but failed

distinct_pandas.ipynb combine the traces, and store the extracted data in result

find_trace.py show the results in video