ICE 10

Link to source code

Team Members:

Name: Avni Mehta, Class Id: 11

Name: (Trinadha) Raji Muppala, Class Id: 14

Task 1:

Train the given dataset for 100 and 200 iterations using the inception model. Provide Tensor Board visualizations report the accuracy.

Input:

Dataset of flowers with 3 classes: sunflower, roses, and tulips.

Code:

We have used the same code provided in the use case. We retrained only the last layer of the GoogleNet with the dataset keeping the initial layers untouched to classify whether our test image is a sunflower, rose, or tulip with scores.

Modified the training steps to 100.

Output:

Accuracy with 100 training steps:

Classification result with 100 training steps:

Accuracy with 200 training steps:

Classification result with 200 training steps:

Tensor board graph:

Tensor board summary:

Tensor board distribution:

Observation:

The test image is identified correctly as rose with the highest score in both cases.

Contritbution of Team Members:

100% for both since both team members completed the assignment individually.