WildCam

import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import matplotlib.pyplot as plt
from PIL import Image, ImageDraw
import json
import cv2
import seaborn as sns
from datetime import datetime
import os
%matplotlib inline

# Input data files are available in the "../input/" directory.
# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory

# loading datasets as json
with open("../input/iwildcam-2020-fgvc7/iwildcam2020_megadetector_results.json") as json_file:
    megadetector_results = json.load(json_file)
with open("../input/iwildcam-2020-fgvc7/iwildcam2020_test_information.json") as json_file:
    test_information = json.load(json_file)
with open("../input/iwildcam-2020-fgvc7/iwildcam2020_train_annotations.json") as json_file:
    train_annotations = json.load(json_file)
    
train_annotations.keys()
# converting into dataframes
train_annot = pd.DataFrame(train_annotations["annotations"])
train_images = pd.DataFrame(train_annotations["images"])
train_cat = pd.DataFrame(train_annotations["categories"])

# Test dataset
test_images = pd.DataFrame(test_information["images"])
test_data = test_images[["file_name","id"]].copy()
test_data['Category'] = 0

# Any results you write to the current directory are saved as output.
train_images.rename({"id":"image_id"}, axis = 1, inplace= True)

train = train_images.merge(train_annot, on = "image_id", how = "inner")

train = train.merge(train_cat.drop("count", axis=1).rename({"id":"category_id"}, axis = 1))
train.tail()
tr=train[['file_name','category_id']]
tr.to_csv('tr.csv',index=False)

print('Number of train images',len(train_images))
print('Number of test images',len(test_images))

# Viewing images in train folder
w, ax = plt.subplots(2,2)
im1 = Image.open('../input/iwildcam-2020-fgvc7/train/939e25fc-21bc-11ea-a13a-137349068a90.jpg')
im2 = Image.open('../input/iwildcam-2020-fgvc7/test/867665c4-21bc-11ea-a13a-137349068a90.jpg')
im3 = Image.open('../input/iwildcam-2020-fgvc7/train/8676382e-21bc-11ea-a13a-137349068a90.jpg')
im4 = Image.open('../input/iwildcam-2020-fgvc7/train/86760c00-21bc-11ea-a13a-137349068a90.jpg')

ax[0,0].imshow(im1)
ax[0,1].imshow(im2) 
ax[1,0].imshow(im3)
ax[1,1].imshow(im4)

tr.head() # csv file for train data
!git clone https://github.com/Tessellate-Imaging/monk_v1.git
! cd monk_v1/installation/Misc && pip install -r requirements_kaggle.txt

# Monk
#Using Mx-net
import os
import sys
sys.path.append("/kaggle/working/monk_v1/monk/");
from gluon_prototype import prototype
gtf = prototype(verbose=1);
gtf.Prototype("iWildCam2020", "Using-densenet121");
gtf.Default(dataset_path="../input/iwildcam-2020-fgvc7/train/",
            path_to_csv="/kaggle/working/tr.csv", # updated csv file 
            model_name="densenet121", 
            freeze_base_network=False,
            num_epochs=1);

gtf.Dataset_Percent(3); # for taking only 3% of dataset for training
gtf = prototype(verbose=1);
gtf.Prototype("iWildCam2020", "Using-densenet121");
gtf.Default(dataset_path="../input/iwildcam-2020-fgvc7/train/",
            path_to_csv="/kaggle/working/sampled_dataset_train.csv", # updated csv file 
            model_name="densenet121", 
            freeze_base_network=False,
            num_epochs=6); 

gtf.EDA(check_corrupt=True); # for checking for corrupt images in 3% dataset

gtf.Train();


gtf = prototype(verbose=1);
gtf.Prototype("iWildCam2020", "Using-densenet121", eval_infer=True);

sample = "../input/iwildcam-2020-fgvc7/test/86767820-21bc-11ea-a13a-137349068a90.jpg";
predictions = gtf.Infer(img_name=sample);
#Display 
from IPython.display import Image
Image(filename=sample)