TallyQA: Answering Complex Counting Questions

Authors: Manoj Acharya, Kushal Kafle, Christopher Kanan8076-8084

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we describe a series of experiments to evaluate the efficacy of multiple algorithms on both simple and complex counting questions. Results for all methods on How Many-QA and both of Tally QA s test sets are given in Table 3.
Researcher Affiliation Academia Manoj Acharya, Kushal Kafle, Christopher Kanan Chester F. Carlson Center for Imaging Science Rochester Institute of Technology {ma7583, kk6055, kanan}@rit.edu
Pseudocode No No structured pseudocode or algorithm blocks were found in the paper.
Open Source Code No No explicit statement or link regarding the open-sourcing of the research's code was found.
Open Datasets Yes We created Tally QA, the world s largest dataset for open-ended counting... It is now publicly available. Tally QA s images are drawn from both COCO and Visual Genome.
Dataset Splits Yes Tally QA is split into one training split (Train) and two test splits: Test-Simple and Test-Complex. Using our simple-complex classifier, Train was found to have 188,439 simple and 60,879 complex questions. The number of questions in each split is given in Table 2.
Hardware Specification No The lab thanks NVIDIA for the donation of a GPU, but no specific GPU model or other hardware specifications for the experiments are provided.
Software Dependencies No The paper mentions software like Faster R-CNN, ResNet-101, Adam optimizer, and SpaCy but does not provide specific version numbers for these or other software dependencies.
Experiment Setup Yes The GRU used in all models are regularized using a dropout of 0.3. RCN is trained using the Adam optimizer with a learning rate of 7e 4 and a batch size of 64 samples.