Applications Algorithms
Automotive application software to optimally combine with Journey Automotive processors
Horizon Engineers Win Four Waymo Challenges In June 2020
Competing with
>100 researchers worldwide
Horizon Achieved
4 World Championships out of 5
Other Participants Include
Waymo open dataset Challenge June 2020
Vision input at 30fps
Champion
Latency
Accuracy
Horizon Engineers Win Waymo Challenge Again in June 2021
Horizon has once again proved its leading algorithm ability with its outstanding results
Under constraints of
< 70ms latency and > 70 APH / L2 (accuracy)
Horizon Won
The 2 main World Championships
Highest accuracy 3D detection model under latency constraint (AFDetv2)
Most efficient model with lowest latency under accuracy constraints (AFDetv2-Base)
55.9ms is the fastest with accuracy constraints of APH/L2 above 70.
#1,2,3 were 2020 candidate without latency constraints
Waymo Dataset Result for June 2021
LiDAR Perception
LiDAR perception models run very efficiently on Journey BPU and can be fused with multi-camera perception
Bird’s Eye View (BEV), A New Perception Paradigme
Need rules to transform the representation into 3D space
Incomplete intermediate results. Could miss important perception when objects stretches across cameras
Per-sensor processing requires late fusion
Post processing is not learnable. Corner cases require large engineering effort
Designed for the optimum runtime execution of Vision and LiDAR perception tasks
Outperforms competition in benchmarks of modern neural networks such as Mobilenetv2 and EfficientNet Bx in latency and throughput
Effectively enables your deep learning model to be optimized for the BPU CNN engine
Complies with automotive quality, reliability and safety standards, including AEC-Q100 grade 2
Horizon Advanced Learned Based Planning
Advanced Planning addresses complex scenario and improves the driving experience
In AD, the Planning stage consumes the BEV perception representation to support automated driving functions
Rule based explicit planning has inherent limitations: Many possible options require to restrict ODDs and enact arbitrary rules Low data utilization Conservative. Not human-like behavior
Imitation planning learns from expert. Driver data Interactive planning learns from interaction and road experiences and takes multi-agent prediction into account
Solutions Served
Journey automotive processors and Matrix solutions are scalable from ADAS to Robotaxi applications as well as in-cabin systems and Matrix solutions
