Opportunities for Collaboration — Students

NJIT STUDENTS! 3D modeling/rendering for species ID and monitoring

Below are two projects in urgent need of development by talented students, suitable for MS students or advanced undergraduates. Both would look good on a CV, could possibly be done for credit, and there is an opportunity to get paid. Plus, I think they are intrinsically cool. If you are interested, please read the notes that follow the descriptions.

1) Efficient development of textured 3D models representing a complete fish assemblage

This project is to begin the development of a set of textured 3D models of coral reef fish species, perhaps with simple articulation, that will be used for model-matching. The required set of species will be specified. The student might apply any of the following approaches: 1) Develop models from scratch based on images and anatomical specifications; 2) Find and, if necessary, adapt existing models available online (many fish species can be adapted into others by straightforward transformations); 3) Start with highly-detailed surface meshes generated from MRI scans of the species in question (for some species only — these meshes will likely need to be simplified and cleaned up). A student taking on this project would need to have good 3D modeling skills, including texture wrapping. Choice of software is open, as long as the final models can be exported to a standard format. The final product would consist of the models plus a write-up about the most efficient method for generating them (especially if the work fulfills a coursework requirement).

2) Hardware-accelerated 3D model matching for fish

This project takes 3D models such as those developed in (1) above, and attempts to find a best-fit orientation/pose match to a segmented 2D image of a fish. Speed is essential, so the idea is to use 3D hardware acceleration to render the model in a wide variety of orientations and poses at high speed, and then send the renderings, ideally directly from the hardware's frame buffer, to a routine that does a fast comparison, for example by comparing a subset of pixels. This last step might also take advantage of hardware acceleration, for example by being encoded in OpenCL. The exact comparison method is open, and choosing a good method would be part of the project. A student taking on this part of the project would need to have existing skills in 3D programming (e.g., OpenGL), and ideally also in (or be willing to learn) OpenCL. The final product would consist of the annotated code, plus a write-up of the methods employed (which, if the approach works, should be publishable).

Notes

The project director is an ecologist, not a computer scientist, and will provide guidance as to the goals of each project, as well as possibly crazy ideas about how to get there. Students must have an appropriate skill-set, and will need to be self-motivated in solving detailed programming and other computer-related puzzles using existing recources (faculty in CS, friends, online forums, etc.). Please don't apply with something like "I know Visual Basic and Javascript, so I'm pretty sure I can do this too."

Each project should take no more than about a semester, working on it part-time (say, 10 hours per week). Hours are negotiable. Getting credit may also be possible, depending on the details of your program. We can approach your Department.

The two projects, combined, represent a proof-of-concept for an ID engine to part of an automated, underwater fish-monitoring insrument. The hardware portion of the instrument is already in the protoype phase, and existing species ID is accomplished by more conventional 2D feature extraction methods. If the 3D approach works even fairly well, its further development will be written into future grant proposals. There will therefore be an opportunity to remain a part of the project, perhaps as a PhD student.

Inquiries should be sent to Gareth Russell (russell@njit.edu). Include the project you are interested in, and a CV that demonstrates your past experience (projects, coursework) in a relevant field.

Opportunities for Collaboration — General

This page is intented to outline various areas where we see opportunities for interesting (and fundable!) research that is cross-disciplinary in nature. If you are interested in any of these areas, please get in touch!

Generally, there seems to be a huge range of opportunities in the application of emerging technologies from the hard sciences and engineering to the collection of ecological data. Most of the areas below fall into this category. Many of these topics would make great projects for graduate students in Math, Computer Science, or Engineering, or possibly even capstone projects for smart undergraduates.

Markov processes for ecological assembly

If you are skilled in Markov processes, you could really help us out! They are commonly used in ecology to model the sequence of species compositions that any community goes through over time, known as community assembly or succession. The challenge is that ecological data are incredibly noisy, so there are lots of interesting applied questions about how much data must be collected to provide any predictability, or to be able to detect possible interventions (such as species addition, or removal) that might push the assembly process into a new trajectory with a new end state. Also, on what timescale is it reasonable to treat the process as homogenous, and when do we need to consider it inhomogenous.

Tracking large and small organisms

One of the emerging areas in ecologcal data collection is the tracking of organisms as they move around a landscape, from elephants and large birds over many years, to beetles in a field over one season. Such tracking has been impossible, or difficult and expensive, until recently. New technologies such as GPS tracking collars have enabled vast leaps, but they are still very expensive. There is a great opportunity for lower-cost technologies to provide tracking ability. One possiblity is might be RFID, especially the low-cost, no-power-required passive kind. Ideas are welcome, whether for large or small organisms. Personally I'd like to be able to track beetles in a field. No-one has done it before! Note: this is potentially commercializable.