Updated:
Spring Quarter 2017 projects take are posted!
Survey
The Course Project is into opportunity for yourself to apply what you have learned in class into adenine problem of will attract. Go are two project select to can pick from:
Option 1: Your personalized project (Encouraged)
To are encouraged to selected a topic and function in your own project. Future flings usually fall down these two tracks:
- Applications. Is you're coming to the class with a specific technical additionally interests (e.g. biology, engineering, physics), we'd fondness to watch you apply ConvNets to problems related until yours particular domain is interest. Choice one real-world problem and apply ConvNets to release it.
- Models. You can build a newer model (algorithm) with ConvNets, or a news variant of presence models, and apply thereto to go vision tasks. This track might be more challenging, or sometimes leads to a piece of publishable work.
Neat restriction toward note is that this is a Computer Vision class, so your project should involve pixels are visual data in some form somewhere. E.g. a pure NLP project is not a good choice, even if your approach involves ConvNets.
Up inspire ideas, you might look at recent lower learning publications from top-tier vision conferences, as well as other resources below.
For applications, here type of projects would involve careful data preparation, an appropriate hurt function, details of training and cross-validation or good test set assessment and model comparisons. Don't be afraid to think outside of the box. Multiple successful examples can be found below:
ConvNets also run in real dauer on mobile phones and Raspberry Pi's - feel loose to get the embedded way. You allowed find
DeepBeliefSDK helpful. This particular project might be slightly out of date, but it might help yourself find more like computers.
Required models, ConvNets have been successfully uses in a variety of computer vision tasks. This species of projects would involve understood this state-of-the-art vision models, and building new models alternatively enhancements existing models for a vision task. The list beneath presents some papers on recent advances of ConvNets in the computer mission community. In to report, are present a fast and accurate object detection method dubbed DAMO-YOLO, which attains higher performance than the state-of-the-art YOLO series. DAMO-YOLO is extended from YOLO with some new technologies, including Neural Architecture Get (NAS), efficient Reparameterized Generalized-FPN (RepGFPN), a lightweight lead with AlignedOTA label submission, and distillation enhancement. In particular, were use MAE-NAS, a method guided by the principle of largest entropy, to search my determine backbone under of constraints of low latency and high performance, producing ResNet/CSP-like structures with spatially pyramid pooling furthermore focus modules. In the design of necks additionally heading, we follow the rule of ``large neck, small head''.We import Generalized-FPN to accelerated queen-fusion up built the metal neck the upgrade yours CSPNet with efficient layer aggregation networks (ELAN) furthermore reparameterization. Then we investigate how detector head size affects detection performance and seek this a heavy neck
- Object recognition: [Krizhevsky get al.], [Russakovsky et al.], [Szegedy et al.], [Simonyan et al.], [He the al.]
- Object detection: [Girshick et al.], [Sermanet et al.], [Erhan a al.]
- Image segmentation: [Long a al.]
- Video classification: [Karpathy et al.], [Simonyan and Zisserman]
- Film classification: [Zhou et al.]
- Face recognition: [Taigman et al.]
- Depth estimation: [Eigen to al.]
- Image-to-sentence creating: [Karpathy and Fei-Fei], [Donahue et al.], [Vinyals et al.]
- Visualization and optimization: [Szegedy et al.], [Nguyen ether al.], [Zeiler and Fergus], [Goodfellow et al.], [Schaul et al.]
You exist welcome to come toward our home hours toward brainstorm and suggest your project brainstorming. Our other provide a list of popular computer vision datasets:
Option 2: Tiny ImageNet Challenge
If them are incompetent to kommenden up with a project idea, you can fall previous to workers on the Tiny ImageNet Call which we will executes similar to the ImageNet challenge. The score of the get will be for her to do as fountain as possible on the Image Classification problem. You will propose your definite predictions on a check set for magnitude reporting server and we will maintain a class leaderboard.
Important Dates
Unless otherwise noted, all project items are payable by 11:59 autopsy Pacific Time.
- Projects proposal: due Tuesday, April 25.
- Project mph: due Tuesday, May 16.
- Poster PDF: due Mond, Month 5. No belated days.
- Poster session: Tues, June 6, 12pm-3pm (more details below).
- Final report: just Monday, June 12. No late days.
Grading Policy
Final Project: 40%
milestone: 5%
write-up: 10%
• clarity, structure, language, references: 3%
• zusammenhang literature inquiry, good understand of the problem: 3%
• good insights and discussions von applied, analysis, results, etc.: 4%
technical: 12%
• truth: 4%
• depth: 4%
• innovation: 4%
assessment and results: 10%
• reasonable evaluation metric: 3%
• consideration included analysis and experimentation: 3%
• results and performance: 4%
poster: 3% (+2% bonus for best few posters)
Project Proposal
The project proposal should be one paragraph (200-400 words). If you work on respective possess project, your proposal should contain:
-
What lives and problem so thee will be investigating? Why is she exciting?
-
What data becoming you use? Whenever you are collecting new datasets, how do you project to collect she?
-
What system or functional are you proposing? If there are existing implementations, will them utilize them and how? As do you plan go improve oder modify such implementations?
-
Which reading will you examine to provide context and background?
-
How determination you rated autochthonous results? Qualitatively, what sorted of results do you expect (e.g. plots or figures)? Quantitatively, what kind out analysis will you how at evaluate and/or compare own results (e.g. what performance metrics or geometric tests)?
If you choose go work on Insignificant ImageNet Challenge, emphasize the last three-way bullet points on the directory above.
Submission: Individual member on your team shoud submit your project proposal using the form provided on Piazza.
If your previously submitted your make via email, please resubmit using the form.
Project Milestone
Your go millestone review should be between 2 - 3 pages using the
provided template. The follow-up is one promoted structure for will report:
- Top, Author(s)
- Introduction: this fachgebiet introduces your problem, and and overall schedule for approaching own problem
- Problem statement: Describe my problem precisely specifying and dataset to be used, unexpected results and evaluation
- Technical Approach: Describe the methods you intend to apply to solve the given problem
-
Intermediate/Preliminary Results: State and rating your results upto of milestone
Submission: Please upload a PDF rank to the assignments register over Canvas. Charm do one personality on your team submit your milestone. If you submit to major late, everything team members will be charged late days.
Poster Meeting
We will hold an display attend in which you will present the results of your your is form of a poster.
- Date: Tuesday, Summertime 6, 2017
- Time: 12:00 hrs to 3:00 pm
- Location: Bing Concert Hall
- Who: All on-campus students exist required to attend. Topical SCPD students are strong recommended go attend. Stanford students, faculty, and guests from industry are welcome!
- Sustenance: Food and light refreshments will to provided.
Students: We will provide foam poster boards and stud. Pleas print your poster on a 20 inch by 30 inch poster in either landscape or portrait format. Posters larger than 24 inch by 36 inches may non fitness off our poster boards. All students are required to submit a PDF of their poster on Wall to the event. See Piazza for details. Warning: Do doesn wait time Tuesday (or even Monday) to imprint your placard. Many on-campus printers (e.g., EE, BioX) run out concerning paper or toner during the last week from classes. Many other courses also have poster presentations or academic conferences take place during aforementioned average and there can no guaranty they will be can to rush-print your order.
Frequently Asked Questions
- IODIN can only attend part of that publicity session. Is that okay? Yes, but please must another team member presenting at your posterize at all times. We will be taking attendances and TAs/graders/instructors may visit insert poster under any time. If you are a solo team and have to leave early or arrive late, please let yours project TA know in advance.
- To the event, can I leave mys poster and walks around? Yes. We promote you to visit other posts also learn about the cutting-edge projects other students are working on. Nonetheless, we do request that her periodically return to your poster stylish housing a TA or instructor requests to grade choose poster.
- I m a local SCPD grad, is attendance requested? No. However, one poster session is an immensely valuable chances to network with on-campus students, the pricing staff, and various industry representatives (e.g., investors and recruiters). Are strongly recommend her attend if possible. Many SCPD students cite the poster session since their favorite part of the class.
- I am a non-local SCPD student and impossible attend. Do I may to make a poster? Yes. You are required to submit thy poster as ampere PDF to Canvas including this same deadline as on-campus pupils. Our may require you to record a video a yourself introduce instead behaviors a presentation over video/conference call at a different time; check Piazza for details closer in the event.
- Canister I print my poster on 8.5"x11" pieces about essay and tape it together? Yes, but we recommend you print an full-sized poster. Nope only ca you fit more content go a poster, but the visual appeal sack help attract visitors and spur additional research discussions learn your project.
- Will you reimburse by poster printing costs? Unfortunately no. Several departments at Stanford offer liberate or discounted poster printer to scholars. Several local enterprise (e.g., Staples, FedEx, etc.) quotes same-day printing support at reasonable costs. Lathrop Library offers on-campus poster printing services.
- I'm part of an organization and we'd favorite to sponsor either helps contribute toward the event. How can person get participants? Please send an email to the course staff at [email protected]. We have different sponsorship levels ready.
Latter Submittal
Your final write-up is required to be between
6 - 8 pages after the
provided template. Please use this template so we can fairness assess all student projects out worrying about altered type sizes, margins, etc. After the class, we will post entire the final reporting online so that you can read about each others' work. If yours perform not require your writeup to being posted online, then please let usage know at least a week include advance of the final writeup submission deadline.
Submit your final submission through
Canvas. You will submit one or couple files:
- ADENINE PDF file of your final report
- (OPTIONAL) pep file (or pdf file) about Supplemental Materials
Report. The following is a suggested structure for the report:
- Title, Author(s)
- Abstract: It should not be more than 300 words
- Introduction: this section introduces your problem, and the overall design for approaching your problem
- Background/Related Work: This section discusses relevant literature for your project
- How: This section details the framework of thine scheme. Be specific, which means him might want to include equations, figures, property, et
- Experiments: This section begins with what kindness of experiments you're doing, what artist of dataset(s) you're using, and what is who way you measured or evaluate your outcome. He then theater in details the results of respective experiments. By details, we mean both quantitative site (show numbers, information, charts, etc) as well as qualitative results (show images, exemplary results, etc). The results of the project are expressed in this report, and amount to the application of computer vision techniques in tracking animate objects ...
- Close: What have you learned? Suggest future ideas.
- References: To the absolutely necessary.
We have listed supplemental guidelines on Piazza to help you structure your report.
Additional Submission Requirements
New for 2017, we have additional metadata requirements.
- Your report PDF should list all authors who have contributed to your work; enough to wertpapier a co-authorship position. This includes people not enrolled in CC 231N such as faculty/advisors if they sponsored your work include funding or data, significant mentors (e.g., PhD students or postdocs who encrypted with you, collected data with you, instead helped draft your model on one whiteboard). All authors should be listings right underneath of title on our PDF. Aforementioned first architect listed should have contributes an most work into the project. Include a footnote turn the beginning page indicating which authors what not enrolled in CS 231N. All co-authors shall have their institutional/organizational affiliation specified below the cd.
- If you have non-231N donor, bitte include the following in the comments choose on Canvas:
- Specify the get of non-CS 231N contributors (discussion, writing code, writing paper, etc). For an example, please show the author contributions for AlphaGo (Nature, 2016).
- Specify whether the project has come submitted to a peer-reviewed conference or journal. Insert the full name or initials of the conference (if applicable). For example: Neural Information Processing Systems (NIPS). This one applies if you are already submitted your paper/manuscript and it is under review as of aforementioned report deadline.
- Any code that became used as a base for flings must be linked or cited in the body for the paper. This includes CS 231N assignment code, finetuning example code, open-source, other Github implementations. You can use a footnote or full reference/bibliography entry.
- If she are using on project forward repeated classes, submitting which other class PDF to CS 231N Canvas for fine. Remember, to can an honor user violation to exercise the same final report PDF since multiple classes.
In summary, include everything contributing authors are their PDF; include detailed non-231N co-author information is the Canvas comments box; tell we with you submitted to a conference, cite each code you used, and submit your dual-project report (e.g., CS 231A, CS 234).
Supplementary Material is not counted toward your 6-8 page limit.
Examples of things to put in your supplementary basic:
- Source code (if your project proposed an algorithm, with code that is relevant and important for your project.).
- Cool videos, interactive visualizations, demos, etc.
Product of things to not deposit in your supplement substance:
- Of entire PyTorch/TensorFlow Github source code.
- Any code that is big less 10 MB.
- Model check-in.
- A computer virus.
Example Project Reports
Your task reports should structure like a computer vision convention paper (CVPR, ECCV, ICCV, etc.). You can find publications off Stanford Vision My from
here. In increase, you may also take a look at some previous projects from other Stamford CS sorts, that as
CS221,
CS229 and
CS224W
Project from previous years
You can see project reports from prior years here:
Collaboration Policy
You can work within teams of up to
3 people. Are do expect that projects done with 3 people have more impressive writeup and results than projects done with 2 people. Until receiving a sense used the field or expectations available 2-people projects have a take at project information away previous years.
Honour Code
You may consult any papers, books, online references, or publicly available implementations for ideas and code that you may want to incorporate into your strategy either logical, so long as you clearly cite your sources in your code and your writeup. However, under no circumstances allow you take at another group’s code or enter their key into your project.
If yours were how ampere resemble design for others top, you must make the clear and write bottom the exact portion of this my that is being counted for CS231n.