Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
On learning to localize objects with minimal supervision
Authors: Hyun Oh Song, Ross Girshick, Stefanie Jegelka, Julien Mairal, Zaid Harchaoui, Trevor Darrell
ICML 2014 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments demonstrate that the proposed approach provides a 50% relative improvement in mean average precision over the current state-of-the-art on PASCAL VOC 2007 detection. |
| Researcher Affiliation | Academia | Hyun Oh Song EMAIL Ross Girshick EMAIL Stefanie Jegelka EMAIL Julien Mairal EMAIL Zaid Harchaoui EMAIL Trevor Darrell EMAIL |
| Pseudocode | No | The paper describes algorithms and mathematical formulations but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | Source code will be available on the author s website. |
| Open Datasets | Yes | We performed two sets of experiments, one on a multiple instance learning dataset (Andrews et al., 2003) and the other on the PASCAL VOC 2007 data (Everingham et al.). |
| Dataset Splits | Yes | For this experiment, we performed 10 fold cross validation on C and µ. Table 1 shows the experimental results. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | No | We use the recently proposed R-CNN (Girshick et al., 2014) detection framework to compute features on image windows in both cases. Specifically, we use the convolutional neural network (CNN) distributed with De CAF (Donahue et al., 2014), which is trained on the Image Net ILSVRC 2012 dataset (using only image-level annotations). |
| Experiment Setup | Yes | For preprocessing, we centered each feature dimension and ℓ2 normalize the data. For fair comparison with (Andrews et al., 2003), we use the same initialization, where the initial weight vector is obtained by training an SVM with all the negative instances and bag-averaged positive instances. |