-
AnyMAL: An Efficient and Scalable Any-Modality Augmented Language Model
Authors:
Seungwhan Moon,
Andrea Madotto,
Zhaojiang Lin,
Tushar Nagarajan,
Matt Smith,
Shashank Jain,
Chun-Fu Yeh,
Prakash Murugesan,
Peyman Heidari,
Yue Liu,
Kavya Srinet,
Babak Damavandi,
Anuj Kumar
Abstract:
We present Any-Modality Augmented Language Model (AnyMAL), a unified model that reasons over diverse input modality signals (i.e. text, image, video, audio, IMU motion sensor), and generates textual responses. AnyMAL inherits the powerful text-based reasoning abilities of the state-of-the-art LLMs including LLaMA-2 (70B), and converts modality-specific signals to the joint textual space through a…
▽ More
We present Any-Modality Augmented Language Model (AnyMAL), a unified model that reasons over diverse input modality signals (i.e. text, image, video, audio, IMU motion sensor), and generates textual responses. AnyMAL inherits the powerful text-based reasoning abilities of the state-of-the-art LLMs including LLaMA-2 (70B), and converts modality-specific signals to the joint textual space through a pre-trained aligner module. To further strengthen the multimodal LLM's capabilities, we fine-tune the model with a multimodal instruction set manually collected to cover diverse topics and tasks beyond simple QAs. We conduct comprehensive empirical analysis comprising both human and automatic evaluations, and demonstrate state-of-the-art performance on various multimodal tasks.
△ Less
Submitted 27 September, 2023;
originally announced September 2023.
-
An AI-Powered VVPAT Counter for Elections in India
Authors:
Prasath Murugesan,
Shamshu Dharwez Saganvali
Abstract:
The Election Commission of India has introduced Voter Verified Paper Audit Trail since 2019. This mechanism has increased voter confidence at the time of casting the votes. However, physical verification of the VVPATs against the party level counts from the EVMs is done only in 5 (randomly selected) machines per constituency. The time required to conduct physical verification becomes a bottleneck…
▽ More
The Election Commission of India has introduced Voter Verified Paper Audit Trail since 2019. This mechanism has increased voter confidence at the time of casting the votes. However, physical verification of the VVPATs against the party level counts from the EVMs is done only in 5 (randomly selected) machines per constituency. The time required to conduct physical verification becomes a bottleneck in scaling this activity for 100% of machines in all constituencies. We proposed an automated counter powered by image processing and machine learning algorithms to speed up the process and address this issue.
△ Less
Submitted 9 December, 2022;
originally announced December 2022.
-
Vehicle Route Planning using Dynamically Weighted Dijkstra's Algorithm with Traffic Prediction
Authors:
Piyush Udhan,
Akhilesh Ganeshkar,
Poobigan Murugesan,
Abhishek Raj Permani,
Sameep Sanjeeva,
Parth Deshpande
Abstract:
Traditional vehicle routing algorithms do not consider the changing nature of traffic. While implementations of Dijkstra's algorithm with varying weights exist, the weights are often changed after the outcome of algorithm is executed, which may not always result in the optimal route being chosen. Hence, this paper proposes a novel vehicle routing algorithm that improves upon Dijkstra's algorithm u…
▽ More
Traditional vehicle routing algorithms do not consider the changing nature of traffic. While implementations of Dijkstra's algorithm with varying weights exist, the weights are often changed after the outcome of algorithm is executed, which may not always result in the optimal route being chosen. Hence, this paper proposes a novel vehicle routing algorithm that improves upon Dijkstra's algorithm using a traffic prediction model based on the traffic flow in a road network. Here, Dijkstra's algorithm is adapted to be dynamic and time dependent using traffic flow theory principles during the planning stage itself. The model provides predicted traffic parameters and travel time across each edge of the road network at every time instant, leading to better routing results. The dynamic algorithm proposed here predicts changes in traffic conditions at each time step of planning to give the optimal forward-looking path. The proposed algorithm is verified by comparing it with conventional Dijkstra's algorithm on a graph with randomly simulated traffic, and is shown to predict the optimal route better with continuously changing traffic.
△ Less
Submitted 30 May, 2022;
originally announced May 2022.
