Visual question answering (VQA) is a machine studying process that requires a mannequin to reply a query about a picture or a set of images. Standard VQA approaches want a considerable amount of labeled coaching information consisting of hundreds of human-annotated question-answer pairs related to pictures. In recent times, advances in large-scale pre-training have led to the event of VQA strategies that carry out properly with fewer than fifty training examples (few-shot) and without any human-annotated VQA training data (zero-shot). Nonetheless, there’s nonetheless a big efficiency hole between these strategies and state-of-the-art absolutely supervised VQA strategies, resembling MaMMUT and VinVL. Specifically, few-shot strategies battle with spatial reasoning, counting, and multi-hop reasoning. Moreover, few-shot strategies have usually been restricted to answering questions on single pictures.
To enhance accuracy on VQA examples that contain complicated reasoning, in “Modular Visual Question Answering via Code Generation,” to look at ACL 2023, we introduce CodeVQA, a framework that solutions visible questions utilizing program synthesis. Particularly, when given a query about a picture or set of pictures, CodeVQA generates a Python program (code) with easy visible capabilities that permit it to course of pictures, and executes this program to find out the reply. We show that within the few-shot setting, CodeVQA outperforms prior work by roughly 3% on the COVR dataset and a couple of% on the GQA dataset.
CodeVQA
The CodeVQA method makes use of a code-writing giant language mannequin (LLM), resembling PALM, to generate Python packages (code). We information the LLM to accurately use visible capabilities by crafting a immediate consisting of an outline of those capabilities and fewer than fifteen “in-context” examples of visible questions paired with the related Python code for them. To pick out these examples, we compute embeddings for the enter query and of the entire questions for which we’ve annotated packages (a randomly chosen set of fifty). Then, we choose questions which have the very best similarity to the enter and use them as in-context examples. Given the immediate and query that we wish to reply, the LLM generates a Python program representing that query.
We instantiate the CodeVQA framework utilizing three visible capabilities: (1) question, (2) get_pos, and (3) find_matching_image.
Question, which solutions a query a couple of single picture, is carried out utilizing the few-shot Plug-and-Play VQA (PnP-VQA) technique. PnP-VQA generates captions utilizing BLIP — an image-captioning transformer pre-trained on thousands and thousands of image-caption pairs — and feeds these right into a LLM that outputs the solutions to the query.Get_pos, which is an object localizer that takes an outline of an object as enter and returns its place within the picture, is carried out utilizing GradCAM. Particularly, the outline and the picture are handed by the BLIP joint text-image encoder, which predicts an image-text matching rating. GradCAM takes the gradient of this rating with respect to the picture options to search out the area most related to the textual content.Find_matching_image, which is utilized in multi-image questions to search out the picture that finest matches a given enter phrase, is carried out by utilizing BLIP textual content and picture encoders to compute a textual content embedding for the phrase and a picture embedding for every picture. Then the dot merchandise of the textual content embedding with every picture embedding characterize the relevance of every picture to the phrase, and we decide the picture that maximizes this relevance.
The three capabilities might be carried out utilizing fashions that require little or no annotation (e.g., textual content and image-text pairs collected from the online and a small variety of VQA examples). Moreover, the CodeVQA framework might be simply generalized past these capabilities to others {that a} consumer would possibly implement (e.g., object detection, picture segmentation, or data base retrieval).
 |
Illustration of the CodeVQA technique. First, a big language mannequin generates a Python program (code), which invokes visible capabilities that characterize the query. On this instance, a easy VQA technique (question) is used to reply one a part of the query, and an object localizer (get_pos) is used to search out the positions of the objects talked about. Then this system produces a solution to the unique query by combining the outputs of those capabilities. |
Outcomes
The CodeVQA framework accurately generates and executes Python packages not just for single-image questions, but additionally for multi-image questions. For instance, if given two pictures, every displaying two pandas, a query one would possibly ask is, “Is it true that there are 4 pandas?” On this case, the LLM converts the counting query in regards to the pair of pictures right into a program during which an object depend is obtained for every picture (utilizing the question perform). Then the counts for each pictures are added to compute a complete depend, which is then in comparison with the quantity within the authentic query to yield a sure or no reply.
 |
We consider CodeVQA on three visible reasoning datasets: GQA (single-image), COVR (multi-image), and NLVR2 (multi-image). For GQA, we offer 12 in-context examples to every technique, and for COVR and NLVR2, we offer six in-context examples to every technique. The desk under reveals that CodeVQA improves persistently over the baseline few-shot VQA technique on all three datasets.
| Methodology | GQA | COVR | NLVR2 | ||||||||
| Few-shot PnP-VQA | 46.56 | 49.06 | 63.37 | ||||||||
| CodeVQA | 49.03 | 54.11 | 64.04 |
| Outcomes on the GQA, COVR, and NLVR2 datasets, displaying that CodeVQA persistently improves over few-shot PnP-VQA. The metric is exact-match accuracy, i.e., the proportion of examples during which the anticipated reply precisely matches the ground-truth reply. |
We discover that in GQA, CodeVQA’s accuracy is roughly 30% greater than the baseline on spatial reasoning questions, 4% greater on “and” questions, and three% greater on “or” questions. The third class contains multi-hop questions resembling “Are there salt shakers or skateboards within the image?”, for which the generated program is proven under.
img = open_image("Image13.jpg")
salt_shakers_exist = question(img, "Are there any salt shakers?")
skateboards_exist = question(img, "Are there any skateboards?")
if salt_shakers_exist == "sure" or skateboards_exist == "sure":
reply = "sure"
else:
reply = "no"
In COVR, we discover that CodeVQA’s achieve over the baseline is greater when the variety of enter pictures is bigger, as proven within the desk under. This pattern signifies that breaking the issue down into single-image questions is helpful.
| Variety of pictures | |||||||||||
| Methodology | 1 | 2 | 3 | 4 | 5 | ||||||
| Few-shot PnP-VQA | 91.7 | 51.5 | 48.3 | 47.0 | 46.9 | ||||||
| CodeVQA | 75.0 | 53.3 | 48.7 | 53.2 | 53.4 | ||||||
Conclusion
We current CodeVQA, a framework for few-shot visible query answering that depends on code technology to carry out multi-step visible reasoning. Thrilling instructions for future work embrace increasing the set of modules used and creating an identical framework for visible duties past VQA. We be aware that care needs to be taken when contemplating whether or not to deploy a system resembling CodeVQA, since vision-language fashions like those utilized in our visible capabilities have been shown to exhibit social biases. On the identical time, in comparison with monolithic fashions, CodeVQA affords further interpretability (by the Python program) and controllability (by modifying the prompts or visible capabilities), that are helpful in manufacturing programs.
Acknowledgements
This analysis was a collaboration between UC Berkeley’s Artificial Intelligence Research lab (BAIR) and Google Analysis, and was performed by Sanjay Subramanian, Medhini Narasimhan, Kushal Khangaonkar, Kevin Yang, Arsha Nagrani, Cordelia Schmid, Andy Zeng, Trevor Darrell, and Dan Klein.
