TSNE Visualization

How to run?

In order to visualize your NxM data, you will need to convert it to CSV file. The format for CSV should follow the following rules.

1. First column: MUST contain image file name.
2. Second column: If you’ve label information for your data points, then add that to the second column of the CSV file. It must be strictly integer value representing the class label. (You may skip this step if there isn’t any class label or just append 0 in entire column)
3. Rest of the columns: Must represent your data points.

Next, you should place your image files in web/img/.

Then, run the code python run.py --target data.csv

Command line arguments

--header	If your CSV file contains header information (in the first row)
--server	Start the server (localhost:8000) to host the TSNE files
--targets	If you have target column within your input CSV file
--max_num_points	Max number of data points to use for visualization

Sample code to construct CSV file

import numpy as np
import pandas as pd
# Lets say you've data in `numpy` array
# 1000 data points of 50D
arr = np.random.rand(1000, 50)

# labels for above data is here
labels = np.random.randint(0, 10, (1000,))
# first convert it to pandas frame
df = pd.DataFrame(arr)

# insert image locations
df.insert(0, 'image', ['image.png']*1000)

# insert targets
df.insert(1, 'target', labels)

# now convert that to CSV
df.to_csv('data.csv', index=False, header=False)

# now you can run the script as follows
# python run.py --target data.csv --server
# and once finished open localhost:8000