Egocentric Video Search via Physical Interactions
Authors: Taiki Miyanishi, Jun-ichiro Hirayama, Quan Kong, Takuya Maekawa, Hiroki Moriya, Takayuki Suyama
AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our proposed method on motion and egocentric video datasets about daily activities in household settings and demonstrate that our egocentric video retrieval framework robustly improves retrieval performance when retrieving past videos from personal and even other persons video archives. |
| Researcher Affiliation | Collaboration | ATR Brain Information Communication Research Laboratory Group, Kyoto, Japan Graduate School of Information Science and Technology, Osaka University, Osaka, Japan {miyanishi, hirayama, kong, t.maekawa, moriyah, suyama}@atr.jp |
| Pseudocode | No | The paper describes its methods using mathematical formulas and prose, but it does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks, nor does it present structured steps in a code-like format. |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for their method is publicly available. |
| Open Datasets | No | We built a dataset by collecting the daily activities of eight subjects (not the researchers) in a house... To make daily activity datasets for retrieval evaluation, we speciļ¬ed 20 activities and their start and end points to the collected sensor dataset. The paper describes the creation of a custom dataset but provides no information about its public availability, a link, DOI, or citation for access. |
| Dataset Splits | Yes | For CCA and PCCA, we tuned the parameter as the number of dimensions d in the latent space learned by CCA and PCCA. For KDE , KDE (CCA) KDE (PCCA) , we tuned the feedback motion and video pairs m. We used leave-one-subject-out cross-validation to tune these parameters among candidates d = {50, 100, 150, 200, 250, 300} and m = {1, 2, 4, 8, 16, 32}, which are optimized for the best performance of the average precision on the validation data of seven subjects (without involving a target subject), and tested it with the target subject dataset. |
| Hardware Specification | No | The paper mentions 'wearable motion sensors, LP-WS1101' and 'a wearable camera, Panasonic HX-A100' used for data collection. However, it does not specify the hardware (e.g., CPU, GPU, memory) used to run the computational experiments or train the models. |
| Software Dependencies | No | The paper mentions 'Caffe' as a deep learning framework and 'VGG' for feature extraction, but it does not provide specific version numbers for these or any other software dependencies used in the experiments. |
| Experiment Setup | Yes | We used leave-one-subject-out cross-validation to tune these parameters among candidates d = {50, 100, 150, 200, 250, 300} and m = {1, 2, 4, 8, 16, 32}, which are optimized for the best performance of the average precision on the validation data of seven subjects (without involving a target subject), and tested it with the target subject dataset. |