Index
PreviousA Few Useful Things to Know About Machine LearningNextScaling Video Analytics on Constrained Edge Nodes
Last updated
Was this helpful?
Last updated
Was this helpful?
- Chen et al., SenSys '15
Cache the results to hide the network delivery and server processing latency
Only send the frames that are largely different from the previous frames
- LiKamWa et al., MobiSys '15
Track identical library calls and reuse computed results across multiple applications
- Zhang et al., MobiCom '15
Group cameras monitoring the same area into clusters
Uploads frames with high “utility”(e.g., object count)
Uploads frames that are different from previous frames(e.g., different object counts)
- Han et al., MobiSys '16
Adaptively pick the best specialized model
- Lu et al., SOCC '16
- Mathut et al., MobiSys '17
A system that can run multiple cloud-scale DL models locally on wearable devices
Interleave the loading of memory-intensive FC layers and the execution of compute-intensive convolution layers
- Shen et al., CVPR '17
Leverage the short-term class skew using model cascade
Train specialized video online
- Kang et al., VLDB '17
Model cascade: difference detector(MSE) → cheap/specialized model → full model
- Zhang et al., NSDI '17
Objective: support efficient real-time analytics for multiple queries which have different quality and lag goals
Offline Phase: use profiler to get a set of pareto-optimal configurations(a combination of knobs) from resource-quality space (with a variant of greedy hill-climbing)
Online Phase: periodically change running queries’ configurations/placement/resource allocation to maximize total utility(quality + lag goals)
- Drolia et al., ICDCS '17
Use edge server as a cache with compute resources(similar to CDN)
- Yi et al., SEC '17
- Kang et al., ASPLOS '17 []
Observed that 1) data transfer latency is often higher than mobile computation latency, especially on wireless networks. 2) inside a model, data size is decreasing at the front-end whereas per-layer latency is higher at the back-end.
NOTE: 2) isn't necessarily true for recent networks with global average pooling
- Poms et al., SIGGRAPH '18
Store videos as tables which are optimized for frame sampling on compressed videos
Express frame operations as dataflow graphs
- Jiang et al., ATC '18
Transfer learning → execute common layers only once
Processing more frames with shared DNN vs. greater per-frame accuracy with specialized DNN
- Jiang et al., SIGCOMM '18
Resource-accuracy tradeoff is affected by some persistent characteristics, so we can reuse configurations over time → temporal correlation
Video cameras with the same characteristics share the same best configurations → cross-camera correlations
Configuration knobs independently impact accuracy → reduce search space
Divide cameras into groups → periodically re-profile “leader” videos
- Zhang et al., SIGCOMM '18
Objective: low latency and high accuracy stream processing in WAN
Ask programmers to write degradation functions and profile those configurations
Adaptively change the configuration at runtime → react to congestions
- Hsieh et al., OSDI '18
Enable low-latency and low-cost querying over large historical video datasets.
At ingest time: classify objects using a cheap CNN, cluster similar objects(KNN search), and index each cluster using top-K most confident classification results.
At query-time: looks up the ingest index for cluster centroids that match the class and classifies them using expensive CNN.
- Liu et al., MobiSys '18
Adaptively select DNN compression techniques based on user demand(Acc/Storage/Comp cost/Latency/Energy)
- Ao et al., SoCC '18
Extend the idea of - enable users to build more complex pipelines
novel straggler mitigation strategy
- Guo et al., ASPLOS '18
- Jain et al., arXiv' 18
- Krishnan et al., CIDR '19
Objective: Indexing and query optimization for VDMS(For complex queries like join)
A novel model for encoding, indexing and storing lineage
- Xu et al., EuroSys '19
- Jiang et al., HotEdgeVideo '19
Proposes a new “camera cluster” abstraction
Saving computing resource
Resource Pooling
Improving analytics quality
Hiding low-level intricacies
- Emmons et al., HotEdgeVideo '19
Split-brain inference
- Canel et al., SysML '19
Assumption: relevant events are rare.
Filter frames by using a micro, binary classifier that extract feature maps from base DNN
- Chin et al., SysML '19
Down-sampling images are sometimes beneficial in terms of accuracy(e.g., removing background noise)
Adaptively scaling video to improve both speed and accuracy of object detectors
- Wang et al., HotCloud 19
Use super-resolution to enhance video quality before running analytics(related: )
- Liu et al., MobiCom '19
Dynamic RoI Encoding: decrease the encoding quality of uninterested areas(use the last processed frame as heuristic)
(Dependency-aware) Parallel streaming and inference: divide frames into slices and parallelize the processing between slices
- Jain et al., HotMobile '19
Leverage cross-camera correlations to reduce resource usage and achieve higher inference accuracy
- Haynes et al., SIGMOD '19
An auto-generated benchmark that evaluates the performance of VDBMS
Let users place an arbitrary number of cameras, each with configurable position, resolution, and field of view
Composite queries and automatically generated ground truth labels
- Fu et al., arXiv '19
- Kang et al., VLDB '20
Objective: Support (approximate) aggregate and limit queries over large video dataset
At ingest time, run object detection on small samples of frames and store them
For each query, use them to train a query-specific proxy model
- Bastani et al., SIGMOD '20
- Du et al., SIGCOMM '20
Iterative video processing driven by server-side DNN
- Li et al., SIGCOMM '20
Dynamically adapts filtering decisions based on feature type, threshold, etc.
- Poddar - et al., Security '20
- Zhang et al., VLDB '20
A system that let users generalize to unbounded vocabularies without manual retraining
- Khani et al., arXiv '20