Multi-Stream Deep Similarity Learning Networks for Visual Tracking
Authors: Kunpeng Li, Yu Kong, Yun Fu
IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate that our tracker (MDSLT) achieves state-of-the-art performance in the visual tracking benchmark compared with recent real-time-speed trackers. |
| Researcher Affiliation | Academia | Kunpeng Li1, Yu Kong1 and Yun Fu1,2 1Department of Electrical and Computer Engineering, Northeastern University, Boston, MA 02115, USA 2College of Computer and Information Science, Northeastern University, Boston, MA 02115, USA {kunpengli,yukong,yunfu}@ece.neu.edu |
| Pseudocode | No | The paper describes the network architecture and loss function mathematically but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any statements about releasing open-source code for the described methodology or links to code repositories. |
| Open Datasets | Yes | Image Net Video dataset from [Russakovsky et al., 2015] has been introduced to the tracking community recently. It includes around 4400 videos with more than one million frames annotated by bounding boxes on the target object. [...] To evaluate the tracking performance, we test and analyze our MDSLT method on a large benchmark dataset (OTB-100) [Wu et al., 2015]. |
| Dataset Splits | Yes | Therefore, we use 90% videos in this dataset for training and the rest 10% for validation. |
| Hardware Specification | No | The paper mentions running at '45 fps on a single GPU' but does not specify the model or detailed specifications of the hardware used for experiments. |
| Software Dependencies | No | The paper mentions using CNNs and references AlexNet, but does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | We train our network for 240K iterations with learning rates 0.01 at the beginning and reduce the learning rate by a factor of 10 at every 80K iterations. We use mini-batches of size 8. |