pytorch
Resize exceeds bounds with bicubic interpolation
🐛 Describe the bug
When applying the resize operation on a floating point image (e.g. bounded between 0 and 1) with bicubic interpolation mode, the resulting tensor might contain values that are higher than the maximum value of the input tensor (e.g. slightly below 0 or above 1).
from torchvision.transforms.v2.functional import resize
from torchvision.transforms import InterpolationMode
img = load_image(...) # generic loading function that returns a floating point image between [0., 1.]
img = resize(x, (64, 64), interpolation=InterpolationMode.BICUBIC, antialias=True)
img.min() # might be lower than 0
img.max() # might be larger than 1
Versions
PyTorch version: 2.1.1 Is debug build: False CUDA used to build PyTorch: 11.8 ROCM used to build PyTorch: N/A
OS: Ubuntu 20.04.6 LTS (x86_64) GCC version: (Ubuntu 7.5.0-6ubuntu2) 7.5.0 Clang version: Could not collect CMake version: version 3.26.4 Libc version: glibc-2.31
Python version: 3.11.6 | packaged by conda-forge | (main, Oct 3 2023, 10:40:35) [GCC 12.3.0] (64-bit runtime) Python platform: Linux-5.4.0-169-generic-x86_64-with-glibc2.31 Is CUDA available: True CUDA runtime version: Could not collect CUDA_MODULE_LOADING set to: LAZY GPU models and configuration: GPU 0: NVIDIA A100-PCIE-40GB GPU 1: NVIDIA A100-PCIE-40GB GPU 2: NVIDIA A100-PCIE-40GB GPU 3: NVIDIA A100-PCIE-40GB GPU 4: NVIDIA A100-PCIE-40GB GPU 5: NVIDIA A100-PCIE-40GB GPU 6: NVIDIA A100-PCIE-40GB GPU 7: NVIDIA A100-PCIE-40GB
Nvidia driver version: 525.147.05 cuDNN version: Probably one of the following: /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn.so.8.1.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_adv_infer.so.8.1.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_adv_train.so.8.1.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_infer.so.8.1.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_cnn_train.so.8.1.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_ops_infer.so.8.1.0 /usr/local/cuda-11.1/targets/x86_64-linux/lib/libcudnn_ops_train.so.8.1.0 HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True
CPU: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Byte Order: Little Endian Address sizes: 43 bits physical, 48 bits virtual CPU(s): 128 On-line CPU(s) list: 0-127 Thread(s) per core: 2 Core(s) per socket: 32 Socket(s): 2 NUMA node(s): 2 Vendor ID: AuthenticAMD CPU family: 23 Model: 49 Model name: AMD EPYC 7452 32-Core Processor Stepping: 0 Frequency boost: enabled CPU MHz: 1499.818 CPU max MHz: 2350.0000 CPU min MHz: 1500.0000 BogoMIPS: 4699.92 Virtualization: AMD-V L1d cache: 2 MiB L1i cache: 2 MiB L2 cache: 32 MiB L3 cache: 256 MiB NUMA node0 CPU(s): 0-31,64-95 NUMA node1 CPU(s): 32-63,96-127 Vulnerability Gather data sampling: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Retbleed: Vulnerable Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Mitigation; Retpolines, IBPB conditional, IBRS_FW, STIBP conditional, RSB filling, PBRSB-eIBRS Not affected Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr wbnoinvd arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif umip rdpid overflow_recov succor smca sme sev sev_es
Versions of relevant libraries:
[pip3] numpy==1.26.2
[pip3] numpy==1.26.2
[pip3] pytorch-lightning==2.1.2
[pip3] torch==2.1.1
[pip3] torchaudio==2.1.1
[pip3] torchdiffeq==0.2.3
[pip3] torchmetrics==1.2.1
[pip3] torchvision==0.16.1
[pip3] torchvision==0.16.1
[pip3] triton==2.1.0
[conda] blas 1.0 mkl
[conda] libblas 3.9.0 12_linux64_mkl conda-forge
[conda] libcblas 3.9.0 12_linux64_mkl conda-forge
[conda] liblapack 3.9.0 12_linux64_mkl conda-forge
[conda] mkl 2021.4.0 h06a4308_640
[conda] numpy 1.26.2 pypi_0 pypi
[conda] pytorch 2.1.1 py3.11_cuda11.8_cudnn8.7.0_0 pytorch
[conda] pytorch-cuda 11.8 h7e8668a_5 pytorch
[conda] pytorch-lightning 2.1.2 pyhd8ed1ab_0 conda-forge
[conda] pytorch-mutex 1.0 cuda pytorch
[conda] torchaudio 2.1.1 dev_0
Thanks for the report @Tomsen1410 . It is a known quirk of bicubic mode on floats (it doesn't happen for bilinear mode). There's no good universal solution here: getting values outside of the range of [0, 1] is technically correct, and doesn't happen all the time either so running a .clamp(0, 1) could be a perf hit for a lot of cases. We've decided to leave it like that (at least for now) and let users manually call .clamp() if they want to ensure the range. I'm curious if this had led to issues in your own preproc?
An alternative would be to use uint8 images instead of float. The range would be preserved and it would be much faster (output values would be slightly different due to clipping of intermediate values).
Thanks for the reply. It was indeed slightly inconvenient. I did not expect this behavior and it caused problems with my transform pipeline further down the line. I ended up writing my own class (ResizeWithClamp) to be able to embed it seamlessly in a transforms.Compose pipeline.
Maybe this behavior could be explained as a sidenote in the docs? Or the transforms.Resize class could have an additional parameter?