- talks - lists.cs.princeton.edu

Colloquium Speaker Kira Radinsky, Mon. Feb 10 at 12:30pm
by Emily C. Lawrence 03 Feb '20

03 Feb '20

CS Colloquium Speaker Dr. Kira Radinsky, Diagnostic Robotics Monday, February 10 - 12:30pm Computer Science - Room 105 Host: Elad Hazan [ https://www.cs.princeton.edu/events/25906 | https://www.cs.princeton.edu/events/25906 ] Learning to predict the future in healthcare Our world faces increasingly complex challenges: we destabilized the climate, haven’t beaten all diseases, and haven’t spread the values of democracy and freedom to large parts of the globe, where violence and riots reign supreme. The world must be fixed in our generation - everyone would agree. But in order to take action, build a plan, we need to see the complete picture, and empower decision makers with tools to make those changes. This decade, we have finally reached a critical amount of data to facilitate the creation of such tools. My work is inspired by Mark Twain’s quote, who once said: “The past does not repeat itself, but it rhymes." Although future events have unique circumstances, they typically follow familiar past patterns. Over the past few years, I devoted my life to development of prediction techniques. My system inferred that Cholera outbreaks in land-locked areas are more likely to occur following storms, especially when preceded by a long drought. Another inference is that genocide events tend to occur following events where local opinion makers describe minority groups as pests. These types of patterns are composed of several abstractions, over variable-term temporal extents and selected from a large number of possible causes. The algorithms I developed deal with the complexity of discovering such patterns. Large-scale digital histories, social and real-time media, and human web behavior are harvested and augmented with human knowledge mined from the web to afford real-time estimations of likelihoods of future events. Most recently, these algorithms have accurately predicted the first Cholera outbreak reported in Cuba in fifty years. These types of actionable predictions, that enable preventative measures, have drawn the attention of a UN genocide-prevention organization and the Gates foundations and illustrate the vast potential for real impact on the state of humanity. In the last few years I have been focusing on applying similar techniques for the healthcare and Pharma, leveraging large amount of data obtained from both medical records, EMR and other medical research results data in a quest to create an AI system for automated medical research and breakthroughs. Bio: Dr. Kira Radinsky is the chairman and CTO of Diagnostic Robotics, where the most advanced technologies in the field of artificial intelligence are harnessed to make healthcare better, cheaper, and more widely available. Dr. Radinsky has founded SalesPredict , acquired by eBay in 2016 and served as eBay Chief Scientist (IL). She gained international recognition for her work at the Technion and Microsoft Research for developing predictive algorithms that recognized the early warning signs of globally impactful events, as disease epidemics and political unrests. In 2013, she was named one of MIT Technology Review’s 35 Young Innovators Under 35 and in 2015 Forbes included her as "30 Under 30 Rising Stars in Enterprise Tech”. She is a frequent presenter at global tech and industry conferences, including TEDx , Wired, Strata Data Science, Techcrunch and publishes in HBR. Radinsky also serves as a board member in Israel Security Authorities and technology board of HSBC bank. She also holds a visiting professor position at the Technion focusing on the application of predictive data mining in medicine.

1 0

4:30pm Tue Feb 4 Elaine Shi (Cornell) talk in E-Quad B205
by Jennifer Rexford 01 Feb '20

01 Feb '20

> > > > EE SEMINAR SERIES > > Speaker: > Elaine Shi, Cornell University > Title: > Practical Oblivious Computation > Day: > Tuesday, February 4, 2020 > Time: > 4:30 pm > Room: > B205 Engineering Quadrangle > Host: > Prateek Mittal > > Abstract: > Oblivious computation refers to the ability to compute on "encrypted" data, such that neither intermediate results nor the program's runtime behavior reveal anything about secret inputs. Oblivious computation can enable privacy-preserving data mining for sensitive data (e.g., genomic or financial data), and allow businesses and individuals to monetize their data without compromising their privacy. > In this talk, I will present a novel, binary-tree based paradigm for constructing Oblivious RAM (ORAM) schemes. ORAM is a generic and powerful primitive that is central to realizing oblivious computation based on either trusted hardware or cryptographic secure computation. Our new ORAM constructions not only solve a thirty-year open theoretical challenge, but also allow ORAM to evolve from a theoretical primitive to a practical building block. Specifically, we made it possible, for the first time, to implement ORAM atop secure processors as well as secure multi-party computation. > Additionally, I will present programming language techniques that not only accelerate oblivious computation, but also make it accessible to real-life programmers who are not security experts. Based on these algorithmic and programming language advances, we have built new platforms for oblivious computation and developed various demo applications such as common data structures, data mining, graph algorithms, and streaming algorithms. Evaluation results suggest that our work has led to four to five orders of magnitude performance improvement for moderately large data sizes, and has enabled a dramatic reduction in application development effort. > > > Bio: > Elaine Shi is an Associate Professor in Cornell University. Her research interests include security, cryptography, distributed systems, foundations of blockchains, and language-based security. She is a recipient of the Packard Fellowship, the Sloan Fellowship, the ONR YIP award, the NSF CAREER award, the NSA Best Scientific Security paper, and various other best-paper and research awards. Elaine obtained her Ph.D. from Carnegie Mellon University. Before joining Cornell, she was an Assistant Professor in UMD. > > This seminar is supported by Korhammer Lecture Series Funds

1 0

Sotiris Apostolakis will present his Pre FPO "Perspective: A Sensible Approach to Speculative Automatic Parallelization" on Monday, February 10, 2020 at 10am in CS 402.
by Nicki Mahler 31 Jan '20

31 Jan '20

Sotiris Apostolakis will present his Pre FPO "Perspective: A Sensible Approach to Speculative Automatic Parallelization" on Monday, February 10, 2020 at 10am in CS 402. The members of his committee are as follows: David August (adviser/reader), Simone Campanoni (Northwestern University/reader), Andrew Appel (examiner), Brian Kernighan (examiner), and Zachary Kincaid (examiner). Everyone is invited to attend his talk. The talk title and abstract follow below. Abstract: The promise of automatic parallelization, freeing programmers from the error-prone and time-consuming process of making efficient use of parallel processing resources, remains unrealized. For decades, the imprecision of memory analysis limited the applicability of automatic parallelization. The introduction of speculation to automatic parallelization overcame these applicability limitations, but caused profitability problems due to high communication and bookkeeping costs for speculation validation and commit. This talk presents Perspective, a speculative parallelization framewo rk that maintains the applicability of prior speculative systems while minimizing the use of high-overhead speculative techniques. Unlike current approaches that perform analysis and transformations independently and in sequence, Perspective integrates, without loss of modularity, memory analysis, speculative techniques, and enabling transformations. Specifically, Perspective integrates the first speculation-aware dependence analysis framework, new efficient speculative privatization transformations, and a planning phase to select a minimal-cost set of parallelization-enabling transforms. By reducing speculative parallelization overheads in ways not possible with prior parallelization systems, Perspective obtains higher overall program speedup (23.0x for 12 general-purpose C/C++ programs running on a 28-core shared-memory machine) than Privateer (11.5x ), the prior automatic speculative-DOALL system with the highest applicability.

1 0

EE Talk: Tuesday, February 4 in B205 - Elaine Shi, Cornell University
by Emily Lawrence 30 Jan '20

30 Jan '20

Speaker: Elaine Shi, Cornell University Title: Practical Oblivious Computation Day: Tuesday, February 4, 2020 Time: 4:30 pm Room: B205 Engineering Quadrangle Host: Prateek Mittal Abstract: Oblivious computation refers to the ability to compute on "encrypted" data, such that neither intermediate results nor the program's runtime behavior reveal anything about secret inputs. Oblivious computation can enable privacy-preserving data mining for sensitive data (e.g., genomic or financial data), and allow businesses and individuals to monetize their data without compromising their privacy. In this talk, I will present a novel, binary-tree based paradigm for constructing Oblivious RAM (ORAM) schemes. ORAM is a generic and powerful primitive that is central to realizing oblivious computation based on either trusted hardware or cryptographic secure computation. Our new ORAM constructions not only solve a thirty-year open theoretical challenge, but also allow ORAM to evolve from a theoretical primitive to a practical building block. Specifically, we made it possible, for the first time, to implement ORAM atop secure processors as well as secure multi-party computation. Additionally, I will present programming language techniques that not only accelerate oblivious computation, but also make it accessible to real-life programmers who are not security experts. Based on these algorithmic and programming language advances, we have built new platforms for oblivious computation and developed various demo applications such as common data structures, data mining, graph algorithms, and streaming algorithms. Evaluation results suggest that our work has led to four to five orders of magnitude performance improvement for moderately large data sizes, and has enabled a dramatic reduction in application development effort. Bio: Elaine Shi is an Associate Professor in Cornell University. Her research interests include security, cryptography, distributed systems, foundations of blockchains, and language-based security. She is a recipient of the Packard Fellowship, the Sloan Fellowship, the ONR YIP award, the NSF CAREER award, the NSA Best Scientific Security paper, and various other best-paper and research awards. Elaine obtained her Ph.D. from Carnegie Mellon University. Before joining Cornell, she was an Assistant Professor in UMD. This seminar is supported by Korhammer Lecture Series Funds

1 0

Mingzhe Wang will present his General Exam "Structured Matching for Phrase Localization" on Monday, February 3, 2020 at 3pm in CS 302
by Nicki Mahler 24 Jan '20

24 Jan '20

Mingzhe Wang will present his General Exam on Monday, February 3, 2020 at 3pm in CS 302. The members of his committee are as follows : Jia Deng (adviser), Olga Russakovsky, and Danqi Chen. Everyone is invited to attend his talk, and those faculty wishing to remain for the oral exam following are welcome to do so. His abstract and reading list follow below. In this paper we introduce a new approach to phrase localization: grounding phrases in sentences to image regions. We propose a structured matching of phrases and regions that encourages the semantic relations between phrases to agree with the visual relations between regions. We formulate structured matching as a discrete optimization problem and relax it to a linear program. We use neural networks to embed regions and phrases into vectors, which then define the similarities (matching weights) between regions and phrases. We integrate structured matching with neural networks to enable end-to-end training. Experiments on Flickr30K Entities demonstrate the empirical effectiveness of our approach. Textbook: Szeliski, R., 2010. Computer vision: algorithms and applications. Springer Science & Business Media. Papers: Kulkarni, G., Premraj, V., Dhar, S., Li, S., Choi, Y., Berg, A.C. and Berg, T.L., 2011. Baby talk: Understanding and generating image descriptions. In Proceedings of the 24th CVPR. Plummer, B.A., Wang, L., Cervantes, C.M., Caicedo, J.C., Hockenmaier, J. and Lazebnik, S., 2015. Flickr30k entities: Collecting region-to-phrase correspondences for richer image-to-sentence models. In Proceedings of the IEEE international conference on computer vision (pp. 2641-2649). Wang, L., Li, Y. and Lazebnik, S., 2016. Learning deep structure-preserving image-text embeddings. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 5005-5013). Rohrbach, A., Rohrbach, M., Hu, R., Darrell, T. and Schiele, B., 2016, October. Grounding of textual phrases in images by reconstruction. In European Conference on Computer Vision (pp. 817-834). Springer, Cham. Karpathy, A., Joulin, A. and Fei-Fei, L.F., 2014. Deep fragment embeddings for bidirectional image sentence mapping. In Advances in neural information processing systems (pp. 1889-1897). Karpathy, A. and Fei-Fei, L., 2015. Deep visual-semantic alignments for generating image descriptions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 3128-3137). Hu, R., Xu, H., Rohrbach, M., Feng, J., Saenko, K. and Darrell, T., 2016. Natural language object retrieval. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 4555-4564). Klein, B., Lev, G., Sadeh, G. and Wolf, L., 2014. Fisher vectors derived from hybrid gaussian-laplacian mixture models for image annotation. arXiv preprint arXiv:1411.7399. Vinyals, O., Toshev, A., Bengio, S. and Erhan, D., 2015. Show and tell: A neural image caption generator. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 3156-3164). Fukui, A., Park, D.H., Yang, D., Rohrbach, A., Darrell, T. and Rohrbach, M., 2016. Multimodal compact bilinear pooling for visual question answering and visual grounding. arXiv preprint arXiv:1606.01847.

1 0

Ariel Schvartzman will present his Pre FPO "Circumventing Lower Bounds in Mechanism and Tournament Design" on Monday, January 27, 2020 in CS at 3pm in CS 302
by Nicki Mahler 21 Jan '20

21 Jan '20

Ariel Schvartzman will present his Pre FPO "Circumventing Lower Bounds in Mechanism and Tournament Design" on Monday, January 27, 2020 in CS at 3pm in CS 302. The members of his committee are: Matt Weinberg (adviser/reader), Mark Braverman (reader), Gillat Kol (examiner), Ran Raz (examiner) and Elad Hazan (examiner). Everyone is invited to attend his talk. The talk title and abstract follow below: Title: Circumventing Lower Bounds in Mechanism and Tournament Design How should a seller price multiple items in order to maximize their revenue from a single buyer? This question has been key in connecting research done by computer scientists and economists to applications such as online ad auctions and spectrum auctions. It was shown almost 40 years ago that if the buyer only cares for one item it is optimal for the seller to offer the buyer the item at a take-it-or-leave-it price. This elegant solution, however, may be far from optimal even for the case with a single buyer interested in just two items. Optimal auction design for two or more items is intricate for a number of reasons: the auctions may be bizarre, computationally hard to find or simply too complex to present to a bidder. Numerous results suggest that (under reasonable complexity assumptions) efficiently finding optimal auctions, even in some simple settings, is impossible. In my talk, I will present numerous ways in which we try to circumvent these impossibility results in order to recover positive results. These include the study of different multi-item models and beyond-worst case analysis for mechanism design. I will also discuss recent advances in tournament design, a field at the intersection of mechanism design and social choice theory. For problems in this field we get around known impossibility results via approximations in order to find approximately strategy-proof tournament rules.

1 0

Claudia Roberts will present her General Exam "Selectively Contextual Bandits" on Thursday, January 23, 2020 at 2pm in CS 402
by Nicki Mahler 16 Jan '20

16 Jan '20

Claudia Roberts will present her General Exam "Selectively Contextual Bandits" on Thursday, January 23, 2020 at 2pm in CS 402. The members of her committee are as follows: Arvind Narayanan (adviser), Matthew Salganik (SOC), and Barbara Engelhardt Everyone is invited to attend her talk, and those faculty wishing to remain for the oral exam following are welcome to do so. Her abstract and reading list follow below. Personalization is an integral part of most web-service applications and determines which experience to display to each member. A popular algorithmic framework used in industrial personalization systems are contextual bandits, which seek to learn a personalized treatment assignment policy in the presence of treatment effects that vary with the observed contextual features of the members. In order to keep the optimization task tractable, such systems can myopically make independent personalization decisions that can conspire to create a suboptimal experience in the aggregate of the member’s interaction with the web-service. We design a new family of online learning algorithms that benefit from personalization while optimizing the aggregate impact of the many independent decisions. Our approach selectively interpolates between any contextual bandit algorithm and any context-free multi-armed bandit algorithm and leverages the contextual information for a treatment decision only if this information promises significant gains over a decision that does not take it into account. Apart from helping users of personalization systems feel less targeted, simplifying the treatment assignment policy by making it selectively reliant on the context can help improve the rate of learning. We evaluate our approach on several datasets including a video subscription web-service and show the benefits of such a hybrid policy. Reading List : Methods/Techniques (Contextual Bandits) * [ https://papers.nips.cc/paper/4321-an-empirical-evaluation-of-thompson-sampl… | An Empirical Evaluation of Thompson Sampling ] * [ https://arxiv.org/pdf/1003.0146.pdf | A Contextual-Bandit Approach to Personalized News Article Recommendation ] * [ https://web.stanford.edu/~bvr/pubs/TS_Tutorial.pdf | A Tutorial on Thompson Sampling ] * [ https://arxiv.org/pdf/1003.5956 | Unbiased Offline Evaluation of Contextual-bandit-based News Article Recommendation Algorithms ] * [ https://www.microsoft.com/en-us/research/wp-content/uploads/2016/06/ftir-on… | Online Evaluation for Information Retrieval ] Textbooks * [ http://incompleteideas.net/book/the-book-2nd.html | Reinforcement Learning: An Introduction ] (Ch 2, 16.7) * [ http://www.mmds.org/#ver21 | Mining of Massive Datasets ] (Ch 8, 9) * [ http://shop.oreilly.com/product/0636920027393.do | Bandit Algorithms for Website Optimization ] Historical Context * [ http://rr.cs.cmu.edu/aaai.pdf | Foundations and Grand Challenges of Artificial Intelligence ] Motivating * [ https://link.springer.com/content/pdf/10.1023%2FA%3A1007046204478.pdf | The Perpetuation of Subtle Prejudice: Race and Gender Imagery in 1990s Television Advertising ] * [ https://www.researchgate.net/publication/5153026_Getting_too_personal_React… | Getting too personal: Reactance to highly personalized email solicitations ] * [ https://science.sciencemag.org/content/sci/347/6221/509.full.pdf | Privacy and human behavior in the age of information ] * [ https://www.sciencedirect.com/science/article/pii/S0167923610001983 | The personalization privacy paradox: An exploratory study of decision making process for location-aware marketing ] * [ https://www.microsoft.com/en-us/research/uploads/prod/2019/07/mehrotra-2017… | Auditing Search Engines for Differential Satisfaction Across Demographics ]

1 0

Hansen Zhang will present his Pre FPO "Practical High Performance Containment on FPGA" on Wednesday, January 15, 2020 at 2pm in CS 301
by Nicki Mahler 14 Jan '20

14 Jan '20

Hansen Zhang will present his Pre FPO " Practical High Performance Containment on FPGA" on Wednesday, January 15, 2020 at 2pm in CS 301. The members of his committee are as follows: David August (Reader and Examiner), Adrian Caulfield (Microsoft- Reader), Sharad Malik (Examiner), and Andrew Appel (Examiner) Everyone is invited to his talk. The talk title and abstract follows below. Title: Practical High Performance Containment on FPGA Abstract: The security of a computer system ultimately rests on the security of the underlying hardware components. A containment-based security architecture secures systems using a single, simple HW root of trust to contain the malicious effects of untrusted components within the system. However, prior containment-based security architecture implementations have suffered from either unreasonable adoption costs or significant performance slowdowns. In this talk, I will present FPGAGuard, the first containment-based security architecture implementation that is both efficient and practical. FPGAGuard secures a commercial off-the-shelf (COTS) computer system with a commodity PCIe FPGA network card as the root of trust and demonstrates reasonable performance on real programs.

1 1

Jeffrey Helt will present his General Exam "FDR: Addressing the Parallelism Gap in Primary-backup Databases" on Tuesday, January 21, 2020 at 1pm in CS 402.
by Nicki Mahler 14 Jan '20

14 Jan '20

Jeffrey Helt will present his General Exam "FDR: Addressing the Parallelism Gap in Primary-backup Databases" on Tuesday, January 21, 2020 at 1pm in CS 402. The members of his committee are as follows: Wyatt Lloyd (adviser), Michael Freedman and Amit Levy. Everyone is invited to attend his talk, and those faculty wishing to remain for the oral exam following are welcome to do so. His abstract and reading list follow below. Asynchronously replicated primary-backup databases are the beating heart of many large applications. The backups provide availability if the primary fails and may serve read-only transactions to increase throughput for all clients and reduce latency for clients in other geographic regions. However, to retain the correctness of overlying applications despite their read-only transaction returning stale values, backups should appear virtually indistinguishable from the primary. Thus, backups must guarantee monotonic prefix consistency, where they present a progressing sequence of the primary's recent states. But monotonic prefix consistency severely constrains a backup's execution because it must exactly copy (or clone) the commit order resulting from the primary's concurrency control. Existing cloned concurrency control protocols guarantee this order by limiting the backup's parallelism, but as a result, a parallelism gap exists between them and the primary. In this paper, we prove this parallelism gap leads to unbounded replication lag, where changes committed by the primary may take infinitely long to propagate to the backup, for some workloads. Unbounded lag recently caused catastrophic failures (i.e., user data loss) and outages in production systems, and thus, our proof reveals such catastrophes are potentially lurking behind every workload change. We then design FDR, the first cloned concurrency protocol to guarantee both monotonic prefix consistency and bounded replication lag. Further, we provide an implementation of FDR that is backward-compatible with existing production databases, making it possible to thwart these potential failures immediately. We demonstrate formally and experimentally FDR provides bounded replication lag under a variety of workloads. Reading List: Hong et al., KuaFu: Closing the parallelism gap in database replication, 2013. Wang et al., Query fresh: log shipping on steroids, 2017. Tu et al., Speedy transactions in multicore in-memory databases, 2013. Zheng et al., Fast Databases with Fast Durability and Recovery Through Multicore Parallelism, 2014. Corbett et al., Spanner: Google's globally distributed database, 2013. Thomson et al., Fast Distributed Transactions and Strongly Consistent Replication for OLTP Database Systems, 2014. Van Renesse and Schneider, Chain Replication for Supporting High Throughput and Availability, 2004. Terrace and Freedman, Object Storage on CRAQ: High-Throughput Chain Replication for Read-Mostly Workloads, 2009. Lamport, The part-time parliament. 1998. Papadimitriou, Serializability of Concurrent Database Updates, 1979. Herlihy and Wing, Linearizability: A correctness condition for concurrent objects, 1990. Textbooks: Tannenbaum and van Steen, Distributed Systems: Principles and Paradigms, 2017.

1 0

Hei Law will present his General Exam, "CornerNet-Lite: Efficient Keypoint Based Object Detection" on Monday, January 20, 2020 at 3:30pm in CS 401
by Nicki Mahler 13 Jan '20

13 Jan '20

Hei Law will present his General Exam, "CornerNet-Lite: Efficient Keypoint Based Object Detection" on Monday, January 20, 2020 at 3:30pm in CS 401. The members of his committee are as follows: Jia Deng (adviser), Adam Finkelstein, and Olga Russakovsky. Everyone is invited to attend his talk, and those faculty wishing to remain for the oral exam following are welcome to do so. His abstract and reading list follow below. Abstract: We propose CornerNet, a new approach to object detection where we detect an object bounding box as a pair of keypoints, the top-left corner and the bottom-right corner, using a single convolution neural network. By detecting objects as paired keypoints, we eliminate the need for designing a set of anchor boxes commonly used in prior single-stage detectors. In addition to our novel formulation, we introduce corner pooling, a new type of pooling layer that helps the network better localize corners. Experiments show that CornerNet achieves a 42.2% AP on COCO. However, this accuracy comes at high processing cost. We also tackle the problem of efficient keypoint-based object detection and introduce CornerNet-Lite. CornerNet-Lite is a combination of two efficient variants of CornerNet: CornerNet-Saccade, which uses an attention mechanism to eliminate the need for exhaustively processing all pixels of the image, and CornerNet-Squeeze, which introduces a new compact backbone architecture. Together these two variants address the two critical use cases in efficient object detection: improving efficiency without sacrificing accuracy, and improving accuracy at real-time efficiency. CornerNet-Saccade is suitable for offline processing, improving the efficiency of CornerNet by 6.0x and the AP by 1.0% on COCO. CornerNet-Squeeze is suitable for real-time detection, improving both the efficiency and accuracy of the popular real-time detector YOLOv3 (34.4% AP at 34ms for CornerNet-Squeeze compared to 33.0% AP at 39ms for YOLOv3 on COCO). Together these contributions for the first time reveal the potential of keypoint-based detection to be useful for applications requiring processing efficiency Reading List: 1. Szeliski, R. (2010). Computer vision: algorithms and applications. Springer Science & Business Media. 2. Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 580-587). 3. Girshick, R. (2015). Fast r-cnn. In Proceedings of the IEEE international conference on computer vision (pp. 1440-1448). 4. Ren, S., He, K., Girshick, R., & Sun, J. (2015). Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in neural information processing systems (pp. 91-99). 5. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 779-788). 6. Redmon, J., & Farhadi, A. (2017). YOLO9000: better, faster, stronger. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 7263-7271). 7. Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C. Y., & Berg, A. C. (2016, October). Ssd: Single shot multibox detector. In European conference on computer vision (pp. 21-37). Springer, Cham. 8. Lin, T. Y., Goyal, P., Girshick, R., He, K., & Dollár, P. (2017). Focal loss for dense object detection. In Proceedings of the IEEE international conference on computer vision (pp. 2980-2988). 9. Tychsen-Smith, L., & Petersson, L. (2017). Denet: Scalable real-time object detection with directed sparse sampling. In Proceedings of the IEEE International Conference on Computer Vision (pp. 428-436). 10. Newell, A., Yang, K., & Deng, J. (2016, October). Stacked hourglass networks for human pose estimation. In European conference on computer vision (pp. 483-499). Springer, Cham. 11. Newell, A., Huang, Z., & Deng, J. (2017). Associative embedding: End-to-end learning for joint detection and grouping. In Advances in Neural Information Processing Systems (pp. 2277-2287).

1 0