The Scripps Plankton Camera (SPC) system, in all its iterations, generates a huge amount of image data; the system captures thousands to hundreds of thousands of images per day. In the short time the pier system has been deployed, it has revealed episodic blooms of fragile diatom chains, many delicate gelatinous organisms, and a form of parasitism not previously reported in the Pacific. To speed the analysis of data coming from the SPC, automated techniques are being developed to label incoming images.
At left is an image of colonial radiolarian taken by the SPC during a sea test on the R/V Sproul in 2014.
The Jaffe Lab, in collaboration with UCSD electrical engineer Dr. Nuno Vasconcelos, is leveraging recent advance in Machine Learning to label the SPC data set. Using Convolutional Neural Networks, we hope to enable new ecological studies by rapidly annotating the huge number of images coming from the SPC. Concurrent experiments are underway to embed taxonomic information about the plankton directly into the learning framework.
- Principal Investigators: Jules S. Jaffe
- Software Engineers: Paul L. D. Roberts, Eric C. Orenstein
- Graduate Students: Eric C. Orenstein