image_model_dashboard

about

An interactive dashboard displaying data from Red Ventures' image model analysis.:chart_with_upwards_trend::bar_chart::chart_with_downwards_trend:

Each point on the scatter plot represents one image's feature and is plotted according to its click through rate. If you hover over a point that is an image with multiple features on the plot a red dotted line will appear indicating their connection. You can filter the points for the confidence level (Google Vision's API assigns floating numbers for how confident they are that a certain feature is in an image).

The other bar graphs display the average click through rate for each image, the impact of having a feature on the image and the average click through rate of images without that feature.

tech stack

Django Framework (Python)
D3.js for data visualizations

what still needs to be done

Send requests to Jenkins image model build for specified business Request has 2 headers:
1. business (the business that the pictures came from)
2. image source (arbitrary name the business gave the images)
Returns: JSON with each object having an image name and feature attributes. For example, there could be 659 objects with 480 attributes each.
```
{
  "imgcontent.1080x1080-us-c1-10sec-to5000_imgtemp.notes-video":
    {   "aliceblue":0.0,
        "aquamarine":0.0,
        "azure_x":0.0,
        "black_x": 0.0,
        "brown":0.016844444,
        "burlywood":0.0,
    },
   "img...etc"
}
```
Use Django model manager to update Django models according to new data
Updating all the graphs according to filters and new builds Currently the scatter plot is the only chart with the function to update
Adding sparsity filter lower bound n and upper bound m we filter out features that occur in: less than n percent of images and/or more than m percent of images
Reformatting y and x axis on scatter plot to adjust for new data X-Axis should remove categories that don't show any data points

run locally/getting started

In the command line...

1. git clone git@github.com:rv-ebarnard/image_model_dashboard.git

2. cd mysite

3. python3 manage.py runserver

4. Go to http://127.0.0.1:8000/viz