From d983bdfc8e315e27f895fb24c6c20a8d128f17b7 Mon Sep 17 00:00:00 2001 From: "lingcai.wl" Date: Wed, 15 Jun 2022 18:37:40 +0800 Subject: [PATCH] [to #42463204] support Pil.Image for image_captioning Link: https://code.alibaba-inc.com/Ali-MaaS/MaaS-lib/codereview/9049211 --- modelscope/pipelines/multi_modal/image_captioning.py | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/modelscope/pipelines/multi_modal/image_captioning.py b/modelscope/pipelines/multi_modal/image_captioning.py index 91180e23..3e5f49d0 100644 --- a/modelscope/pipelines/multi_modal/image_captioning.py +++ b/modelscope/pipelines/multi_modal/image_captioning.py @@ -84,8 +84,11 @@ class ImageCaptionPipeline(Pipeline): s = torch.cat([s, self.eos_item]) return s - patch_image = self.patch_resize_transform( - load_image(input)).unsqueeze(0) + if isinstance(input, Image.Image): + patch_image = self.patch_resize_transform(input).unsqueeze(0) + else: + patch_image = self.patch_resize_transform( + load_image(input)).unsqueeze(0) patch_mask = torch.tensor([True]) text = 'what does the image describe?' src_text = encode_text(