diff --git a/.github/workflows/docker.yml b/.github/workflows/docker.yml
index 70507e4d05..6ece3f59a8 100644
--- a/.github/workflows/docker.yml
+++ b/.github/workflows/docker.yml
@@ -12,7 +12,7 @@ jobs:
     steps:
       - uses: actions/checkout@v3
       - name: Build docker image
-        run: docker build . -t doctr-tf-py3.8-slim
+        run: docker build -t doctr-tf-py3.8-slim --build-arg SYSTEM=cpu .
       - name: Run docker container
         run: docker run doctr-tf-py3.8-slim python3 -c 'import doctr'
 
diff --git a/.github/workflows/public_docker_images.yml b/.github/workflows/public_docker_images.yml
index f31e5a2539..de77d1a265 100644
--- a/.github/workflows/public_docker_images.yml
+++ b/.github/workflows/public_docker_images.yml
@@ -50,9 +50,6 @@ jobs:
           tags: |
             # used only on schedule event
             type=schedule,pattern={{date 'YYYY-MM'}},prefix=${{ matrix.framework }}-py${{ matrix.python }}-${{ matrix.system }}-
-            # set latest tag only if `enable` is True
-            # see https://github.com/docker/metadata-action#latest-tag
-            type=raw,value=latest,enable=${{ matrix.framework == 'tf' && matrix.python == '3.8.18' && matrix.system == 'gpu' && github.ref == format('refs/heads/{0}', 'main') }}
             # used only if a tag following semver is published
             type=semver,pattern={{raw}},prefix=${{ matrix.framework }}-py${{ matrix.python }}-${{ matrix.system }}-
 
diff --git a/README.md b/README.md
index 912c14d52a..940c4a4a8b 100644
--- a/README.md
+++ b/README.md
@@ -272,24 +272,25 @@ To verify and configure GPU support for Docker, please follow the instructions p
 Once Docker is configured to use GPUs, you can run docTR Docker containers with GPU support:
 
 ```shell
-docker run --it --gpus all ghcr.io/mindee/doctr:tf-py3.8.18-v0.7.1 bash
+docker run --it --gpus all ghcr.io/mindee/doctr:tf-py3.8.18-gpu-2023-09 bash
 ```
 
 #### Available Tags
 
-The Docker images for docTR follow a specific tag nomenclature: `<framework>-py<python_version>-<doctr_version|YYYY-MM>`. Here's a breakdown of the tag structure:
+The Docker images for docTR follow a specific tag nomenclature: `<framework>-py<python_version>-<system>-<doctr_version|YYYY-MM>`. Here's a breakdown of the tag structure:
 
 - `<framework>`: `tf` (TensorFlow) or `torch` (PyTorch).
 - `<python_version>`: `3.8.18`, `3.9.18`, or `3.10.13`.
+- `<system>`: `cpu` or `gpu`
 - `<doctr_version>`: a tag >= `v0.7.1`
-- `<YYYY-MM>`: e.g. `2023-10`
+- `<YYYY-MM>`: e.g. `2023-09`
 
 Here are examples of different image tags:
 
 | Tag                        | Description                                       |
 |----------------------------|---------------------------------------------------|
-| `tf-py3.8.18-v0.7.1`       | TensorFlow version `3.8.18` with docTR `v0.7.1`. |
-| `torch-py3.9.18-2023-10`| PyTorch version `3.9.18` with a monthly build from `2023-10`. |
+| `tf-py3.8.18-cpu-v0.7.1`       | TensorFlow version `3.8.18` with docTR `v0.7.1`. |
+| `torch-py3.9.18-gpu-2023-09`| PyTorch version `3.9.18` with GPU support and a monthly build from `2023-09`. |
 
 #### Building Docker Images Locally