| # SPDX-FileCopyrightText: Copyright 2026 Arm Limited and/or its affiliates <open-source-office@arm.com> | |
| # SPDX-License-Identifier: Apache-2.0 | |
| schema_version: 1.0.0 | |
| report_type: image-depth-estimation | |
| task_type: monocular-depth-estimation | |
| title: ZipDepth FP16 ExecuTorch — Compute Flow | |
| description: >- | |
| Monocular relative-depth package derived from the unfold-free ZipDepth Base mobile/NPU | |
| checkpoint. The package uses a static ExecuTorch depth model with FP32 tensor interfaces, | |
| FP16 activations, FP32 serialized convolution weights, and dynamic ExecuTorch transforms to | |
| preserve the source image shape at its boundaries. | |
| family_name: ZipDepth | |
| id: Arm/zipdepth-fp16-executorch | |
| filename: zipdepth_base_fp16.pte | |
| base_model_id: fabiotosi92/ZipDepth | |
| vendor: University of Bologna | |
| base_model_url: https://github.com/fabiotosi92/ZipDepth | |
| weight_dtype: fp32 | |
| profile: Arm-Optimized | |
| model_size_mb: 24.609984 | |
| parameter_count: 6142058 | |
| format: executorch | |