EECS 207
Digital Image Processing
Bird Call Project
The ultimate goal of this project will be to search for bird calls in audio recordings using content-based image retrieval (CBIR).
The steps will be:
  1. Create spectrogram images of the audio recordings.
  2. Extract regions of interest (ROI) subimages that correspond to the bird calls. These ROIs have been marked by bird experts.
  3. Extract visual features from the ROI subimages.
  4. Given a query ROI subimage, use the visual features to retrieve visually similar subimages. Hopefully, these will correspond to calls from the same bird species.
You will experiment with:
  1. Extracting the "optimal" spectrogram images.
  2. Extracting different visual features.
  3. Different ways of comparing visual features.
Leaderboard:
Date
 Average Precision
 Configuration
5/4/21
 0.9337 Deep features, mel spectrograms, L2 distance
5/12/21
 0.6547 Gabor features (4,6), Minkowski distance with p=3, no feature normalization, spectrogram computed using window size of 1024 and overlap of 512 and log10 normalization
4/29/21
 0.6473
 From report instructions: Gabor features (4,6), L2 distance, no feature normalization, spectrogram computed using window size of 1024 and overlap of 512 and log10 normalization.



2019
 0.7789


4/7/21

Here is the code (Python and Matlab) from the lab session I held on 4/7/21:

Lab_210407.zip

See the readme.txt file in either the Python_Code or Matlab_Code folders.

4/10/21

Here is the bird call data and supporting code (Python and Matlab):

Bird_Data_And_Code.zip

I went over this in lecture on Thursday, April 8. See the Zoom recording.

See in the .zip file:

- readme.txt
- CourseProject_Dataset.pdf

TASK: BY TUESDAY, APRIL 13: Run and understand the code (Python or Matlab) in:

- compute_spectrograms.*
- identify_rois.*

4/19/21

Here is the Python code for simple feature extraction and retrieval that I went over in lab on Wednesday, April 15:

Feature_Extraction_And_Query.zip

See the Zoom recording for details.

4/29/21

Project report instructions.

4/29/21

Here is the Python code for Gabor texture feature extraction and retrieval that I went over in lecture on Thursday, April 29:

Texture_Feature_Extraction_And_Query.zip

See the Zoom recording for details.

5/3/21

Here are my lecture notes for CBIR:

EECS_207_Notes_CBIR_pp_1_5.pdf
EECS_207_Notes_CBIR_pp_6_16.pdf
EECS_207_Notes_CBIR_pp_17_36.pdf