the code use ONLY OpenCV an matplotlib in Python
Instructions
The idea is that you will develop a smart filter to be added to a new app that a start-up
company is launching in September. You will have the freedom to decide your filter
functionality. It can be anything from image enhancement, fun effects or scientific applications.
The ‘SMART’ in your filter is the most important part of your evaluation.
Process:
You have to write a proposal, in which you describe the functionality of your smart filter.
Submit it for approval and once approved you will be able to start working on your code. You
will prepare a 1-minute video demonstration of your filter. If the majority agree, we will run a
competition in which the entire group will try each filter and anonymously rate it on a fivestar
scale.
Details:
In the proposal (maximum 250 words) you must include: 1. The name of your filter (be creative).
2. a description of the process (only major steps) using at least three of the methods we have
reviewed in class (you can use more and also other methods). 3. You need to justify the use
of each step, e.g. applying convolution three times does not count as three steps.
Teams are allowed, but only to distribute the workload for a larger and modular system. e.g.
if your filter needs to perform 9 different major operations, you can create a team of 3, in
which each one will create a module of the system. In this case, each member of the team
needs to submit their part individually (proposal, video and code), each individual proposal of
a team must start with a brief description of the whole system in less than 100 words (this will
be the same for all team members). Note that each member must implement at least 3 major
operations. Hence the workload/report in a team is exactly as if you are working alone.
The code must be Python, you can use any libraries, remember that we will evaluate the
functionality of the filter, not the individual operations. Your code must be able to run on any
computer, either for the competition or just to be evaluated by the TAs. then your
implementation should be easily executable (more details below).
You can submit your proposal at any time to get approval and I recommend that you submit
earlier to get feedback, the deadline for the proposals is July 5th. This date is after the 9th
class, so you have more time to decide what you want to implement. However, consider that
if all of you submit your proposal on the 5th you will have feedback in a couple of days, and
if your proposal is rejected you will need to rewrite it, resubmit it and wait for feedback again,
resulting in a delay of one week to start your implementation reducing your time to the
deadline of the final project.
The video and code comprise your final project and must be submitted before July 26st at
23:59 (ADT) as stated in the syllabus. Include the image(s) used in your demonstration for
testing and use png or mp4 only.
Evaluation:
We will evaluate your project based on:
1.
Clarity of the proposal. The filter has a clear goal and the methodology justifies the use
of the steps selected.
2.
Readiness of your code. We want to be able to read your code easily, see this as an
extension of the methodology in your proposal, use comments to explain the steps.
3.
Functionality. Your filter most works on different images, we will test each filter with the
set of images provided by you but also with new images and if we don’t get the expected
result based on the video demonstration you will get a lower score.
Just as a reminder here is the list of the topics that we covered (or will be covered) in the class,
each one can be used as one step for your filter:
Convolution
Fast Fourier Transform
Subsampling/scaling
Edge detection
Corner detection
Scale Invariant Feature Transform
Depth map (at least two images)
Optical flow (at least two images)
Segmentation
Boundary fitting (model and image)
Image registration (template and image)
Image classification (model and image)
Remember that each step of the list can have different implementations, there are several
algorithms to perform similar goals, so you are free to select the one that works better for
your project. I have repeated this almost in every lecture, the methods and the parameters
that you will use depend on the problem that you are solving. Most of the time you need to
find the solution in a heuristic way, especially when you start working on real-world scenarios.
Additional details :
You must submit to brigthspace/finalProject a single zip file that contains the following:
1. the video demonstration, (mp4) explaining the aim of the filter and how it works.
2. your function filtername.py
3. the sample image/video (png or mp4 only) that you used in your demonstration and
4. a dependency installer if needed.
If you fail to upload any of those 4 items, your score will be zero.
Note that the code is .py, so it can be run on a terminal. your function will receive the
image/video as an argument, for example:
$ python3 myfiltername.py image1.png
Functions using two images will use two arguments
$ python3 myfiltername.py image1.png image2.png
and functions that use options may have different number of arguments
$ python3 myfiltername.py video.mp4 image1.png arg1 arg2
Here arg1 and arg2 can be for example the amount of effect or just a flag to do a different
processing. This depends on your proposal and what we discussed about your particular filter
so it might be different for all of you.
The output should be a new file. So if you are using video you can process the frames and
then create and save a new mp4 file with the result. Some procedures can take time so it is
better to not plot frame by frame in real-time. If you want you could add one argument to
plot, but the default should do the processing and create the new file. You can print out text
in the terminal if you want.
If you want to avoid variable arguments in your function, a GUI will be appreciated.
Finally another reminder: Use different images/videos when trying your code, We will test your
implementations and your score will depend on performance. Just be sure to use different
conditions so your code works well in different situations.
Do not ask for extensions, there is no reason I will give you one. However, you can use one
SDA (if you still have one, remember the deadline is on Sunday). For those that do not have
an SDA, if you are running late you can submit your materials one day after with 25% reduction
or two days after with 50% reduction.
