Phrase-based Image Captioning
Authors: Remi Lebret, Pedro Pinheiro, Ronan Collobert
ICML 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our approach, which is considerably simpler than state-of-the-art models, achieves comparable results in two popular datasets for the task: Flickr30k and the recently proposed Microsoft COCO. |
| Researcher Affiliation | Collaboration | R emi Lebret REMI@LEBRET.CH Pedro O. Pinheiro PEDRO@OPINHEIRO.COM Idiap Research Institute, Martigny, Switzerland Ecole Polytechnique F ed erale de Lausanne (EPFL), Lausanne, Switzerland Ronan Collobert RONAN@COLLOBERT.COM Facebook AI Research, Menlo Park, CA, USA |
| Pseudocode | Yes | Figure 4. The constrained language model for generating description given the predicted phrases for an image. |
| Open Source Code | No | The paper does not provide a direct link to its source code or explicitly state that the code is open-source. |
| Open Datasets | Yes | The Flickr30k dataset contains 31,014 images where 1,014 images are for validation, 1,000 for testing and the rest for training (i.e. 29,000 images). The COCO dataset contains 123,287 images, 82,783 training images and 40,504 validation images. The testing images has not yet been released. We thus use two sets of 5,000 images from the validation images for validation and test, as in Karpathy & Fei-Fei (2015)1. 1Available at http://cs.stanford.edu/people/ karpathy/deepimagesent/. |
| Dataset Splits | Yes | The Flickr30k dataset contains 31,014 images where 1,014 images are for validation, 1,000 for testing and the rest for training (i.e. 29,000 images). The COCO dataset contains 123,287 images, 82,783 training images and 40,504 validation images. The testing images has not yet been released. We thus use two sets of 5,000 images from the validation images for validation and test, as in Karpathy & Fei-Fei (2015)1. |
| Hardware Specification | Yes | It takes 2.5 hours on single CPU (Intel i7 4930K 3.4 GHz) to train on the COCO training dataset. |
| Software Dependencies | No | The paper mentions using SENNA software and VGG CNN, but does not provide specific version numbers for these or other software dependencies. |
| Experiment Setup | Yes | The parameters θ are V R400 4096 (initialized randomly) and U R400 |C| (initialized with the phrase representations) which are tuned on the validation datasets. They are trained with 15 randomly chosen negative samples and a learning rate set to 0.00025. |