Hotels-50K: A Global Hotel Recognition Dataset

Authors: Abby Stylianou, Hong Xuan, Maya Shende, Jonathan Brandt, Richard Souvenir, Robert Pless726-733

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We present a baseline approach based on a standard network architecture and a collection of data-augmentation approaches tuned to this problem domain.In order to set the baseline for performance on the Hotels50K dataset, we compare two off-the-shelf pre-trained networks trained for object and scene recognition to a method using data and augmentation schemes specifically tailored to hotel recognition.Table 4 shows the results for the ablation experiment.
Researcher Affiliation Collaboration 1George Washington University 2Adobe Research 3Temple University
Pseudocode No The paper describes the methods verbally (e.g., "Our method uses the Hotels-50K training set as input to fine tune a Resnet-50 model...") and visually (Figure 8 for data augmentation), but it does not present any structured pseudocode or algorithm blocks.
Open Source Code Yes The Hotels-50K dataset, pre-trained models and code to replicate our baseline approaches can be found at https://github.com/GWUvision/Hotels-50K.
Open Datasets Yes The Hotels-50K dataset, pre-trained models and code to replicate our baseline approaches can be found at https://github.com/GWUvision/Hotels-50K.
Dataset Splits No The paper mentions "The test set consists of 17,954 images from the Traffick Cam mobile application from 5,000 different hotels, which are a subset of those found in the training set." and "Training parameters were selected using cross-validation." However, it does not explicitly detail the size or composition of a distinct validation set or the specific splits for the cross-validation used for parameter selection.
Hardware Specification No The paper describes the models and training process but does not provide any specific hardware details such as GPU models, CPU types, or memory used for the experiments.
Software Dependencies No The paper mentions using "Resnet-50 model, pre-trained for Image Net" and a "VGG model trained on the Places365 dataset", and references a triplet loss paper. It also mentions MS-COCO for masks. However, it does not provide specific version numbers for any software libraries, frameworks, or dependencies used (e.g., Python, PyTorch/TensorFlow, CUDA versions).
Experiment Setup Yes The final model was fine-tuned for 65,000 iterations with 120 images per batch.Training parameters were selected using cross-validation.